{"id": "2603.21583", "title": "HACMatch Semi-Supervised Rotation Regression with Hardness-Aware Curriculum Pseudo Labeling", "abstract": "Regressing 3D rotations of objects from 2D images is a crucial yet challenging task, with broad applications in autonomous driving, virtual reality, and robotic control. Existing rotation regression models often rely on large amounts of labeled data for training or require additional information beyond 2D images, such as point clouds or CAD models. Therefore, exploring semi-supervised rotation regression using only a limited number of labeled 2D images is highly valuable. While recent work FisherMatch introduces semi-supervised learning to rotation regression, it suffers from rigid entropy-based pseudo-label filtering that fails to effectively distinguish between reliable and unreliable unlabeled samples. To address this limitation, we propose a hardness-aware curriculum learning framework that dynamically selects pseudo-labeled samples based on their difficulty, progressing from easy to complex examples. We introduce both multi-stage and adaptive curriculum strategies to replace fixed-threshold filtering with more flexible, hardness-aware mechanisms. Additionally, we present a novel structured data augmentation strategy specifically tailored for rotation estimation, which assembles composite images from augmented patches to introduce feature diversity while preserving critical geometric integrity. Comprehensive experiments on PASCAL3D+ and ObjectNet3D demonstrate that our method outperforms existing supervised and semi-supervised baselines, particularly in low-data regimes, validating the effectiveness of our curriculum learning framework and structured augmentation approach.", "date": "2026-03-23", "categories": []}
{"id": "2606.18429", "title": "CAOA -- Completion-Assisted Object-CAD Alignment", "abstract": "Accurately aligning CAD models to their corresponding objects in indoor RGB-D scans is a central challenge in 3D semantic reconstruction. The task requires estimating a 9-Degree-of-Freedom (DoF) pose-position, rotation, and scale along three axes-but is hindered by noisy and incomplete scans, as well as segmentation errors that cause geometric distortions. We present Completion-Assisted Object-CAD Alignment (CAOA), a method that integrates a semantically and contextually aware point cloud completion module with a symmetry-aware relative pose estimation algorithm, enabling precise alignment of CAD models to scanned objects. Existing completion methods are typically trained and evaluated on synthetic datasets, which often fail to generalize to real-world scans. To bridge this gap, we introduce a synthetic data generation strategy tailored to indoor scenes, significantly reducing the synthetic-to-real domain gap-validated through quantitative comparisons with widely used completion datasets. In addition, we release S2C-Completion, an expert-annotated dataset of over 8,500 object-CAD pairs from Scan2CAD, created for real-world indoor single-object completion and intended as a new benchmark for this task. For object-CAD alignment, we incorporate symmetry information via a symmetry-aware loss, improving robustness to symmetric ambiguities. On the Scan2CAD benchmark, CAOA achieves a 17% accuracy improvement over state-of-the-art methods.", "date": "2026-06-16", "categories": []}
{"id": "2512.10840", "title": "PoseGAM: Robust Unseen Object Pose Estimation via Geometry-Aware Multi-View Reasoning", "abstract": "6D object pose estimation, which predicts the transformation of an object relative to the camera, remains challenging for unseen objects. Existing approaches typically rely on explicitly constructing feature correspondences between the query image and either the object model or template images. In this work, we propose PoseGAM, a geometry-aware multi-view framework that directly predicts object pose from a query image and multiple template images, eliminating the need for explicit matching. Built upon recent multi-view-based foundation model architectures, the method integrates object geometry information through two complementary mechanisms: explicit point-based geometry and learned features from geometry representation networks. In addition, we construct a large-scale synthetic dataset containing more than 190k objects under diverse environmental conditions to enhance robustness and generalization. Extensive evaluations across multiple benchmarks demonstrate our state-of-the-art performance, yielding an average AR improvement of 5.1% over prior methods and achieving up to 17.6% gains on individual datasets, indicating strong generalization to unseen objects. Project page: https://windvchen.github.io/PoseGAM/ .", "date": "2025-12-11", "categories": []}
{"id": "2601.13565", "title": "Learning Fine-Grained Correspondence with Cross-Perspective Perception for Open-Vocabulary 6D Object Pose Estimation", "abstract": "Open-vocabulary 6D object pose estimation empowers robots to manipulate arbitrary unseen objects guided solely by natural language. However, a critical limitation of existing approaches is their reliance on unconstrained global matching strategies. In open-world scenarios, trying to match anchor features against the entire query image space introduces excessive ambiguity, as target features are easily confused with background distractors. To resolve this, we propose Fine-grained Correspondence Pose Estimation (FiCoP), a framework that transitions from noise-prone global matching to spatially-constrained patch-level correspondence. To systematically eliminate background interference, FiCoP first employs an object-centric disentanglement step to isolate the target from macro-level environmental noise. Building upon this localized region, our core methodological innovations are twofold. Firstly, a Cross-Perspective Global Perception (CPGP) module is proposed to fuse dual-view features, establishing structural consensus through explicit context reasoning and text-guided semantic injection. Secondly, we design a Patch Correlation Predictor (PCP) that leverages a patch-to-patch correlation matrix as a structural prior. This generates a precise block-wise association map, acting as a spatial filter to enforce fine-grained, noise-resilient matching. Experiments on the REAL275 and Toyota-Light datasets demonstrate that FiCoP improves Average Recall by 8.0% and 6.1%, respectively, compared to the state-of-the-art method, highlighting its capability to deliver robust and generalized perception for robotic agents operating in complex, unconstrained open-world environments. The source code will be made publicly available at https://github.com/zjjqinyu/FiCoP.", "date": "2026-01-20", "categories": []}
{"id": "2606.11381", "title": "From Simulation to the Real-World: An In-Field 6D Pose Dataset and Baseline for Robotic Strawberry Harvesting", "abstract": "Robotic strawberry harvesting requires precise 6D pose estimation; however, collecting 6D pose ground truth in real agricultural fields is inherently challenging. Existing strawberry 6D pose estimation studies have therefore relied mainly on synthetic data, often without sufficient scene-level realism,leaving their performance under real agricultural field conditions unquantified. In this work, we present, to the best of our knowledge, the first real-world 6D pose ground truth dataset of strawberries collected in actual agricultural fields (12,040 images). We also introduce a synthetic dataset rendered in NVIDIA Isaac Sim, featuring scene-level realism and domain randomization. Despite this improved simulation setup, our experiments reveal that a substantial sim-to-real gap persists, underscoring the necessity of real agricultural field data for reliable evaluation. We further quantify the sim-to-real gap through baseline 6D pose estimation results across backbone encoders, serving as a reference for future work.", "date": "2026-06-09", "categories": []}
{"id": "2606.16593", "title": "Rotational Symmetry based Object Pose Estimation from Point Clouds in the Absence of Known 3D Models", "abstract": "Object pose estimation is crucial to many industrial applications, with one example being automated spray painting using a robot. However, confidentiality concerns often limit access to high-quality 3D models, posing a significant challenge for point-cloud-based pose estimation. In such scenarios, rotational symmetry, a readily accessible characteristic of many industrial objects, can provide valuable prior information to facilitate pose estimation.In this paper, we propose a method that leverages the rotational symmetry commonly found in industrial objects to address the challenge caused by the absence of 3D models. The object pose is jointly estimated with point cloud refinement through an iterative optimization process. This optimization relies on a rotational symmetry constraint loss. To construct this loss, each 3D point is rotated according to the currently estimated pose, and multiple correspondences are identified using nearest-neighbor search by exploiting the rotational symmetry property. These correspondences are then used to compute the rotational symmetry constraint loss, which iteratively refines both the pose and the point cloud.By explicitly incorporating rotational symmetry into the optimization process, the proposed method achieves robust pose estimation and generalizes well across diverse object types. The proposed method is evaluated on a dataset specifically created for point clouds without known 3D models, consisting of four categories of synthetic objects and one real wheel hub collected from a production line. Experimental results demonstrate that the proposed method achieves performance comparable to methods that rely on known 3D models.", "date": "2026-06-15", "categories": []}
{"id": "2606.14389", "title": "MooMIns -- Monocular 3D Reconstruction and Object Pose Estimation from Multiple Instances", "abstract": "Simultaneous 3D reconstruction and 6D object pose estimation from a single monocular image is an inherently ill-posed problem. In industrial settings, however, multiple instances of an object are often randomly arranged in bins, implicitly providing several views of the same object within a single image. We show that this implicit multi-view geometry can be exploited to simultaneously reconstruct the object in 3D and estimate the 6D pose of each visible object instance. We present MooMIns, a new Gaussian-splatting-based approach that inverts the original Gaussian splatting formulation: instead of rendering a single scene from multiple cameras, we render multiple object instances from a single camera. Our method is initialized with SAM3 instance segmentation masks and a modified Structure from Motion (SfM) pipeline. In contrast to learned monocular depth estimation, we perform true geometry-based reconstruction from image evidence, avoiding hallucinations caused by training data priors. We evaluate MooMIns on synthetic and real bin-picking scenarios, and demonstrate accurate reconstruction of previously unseen objects as well as reliable pose estimation of individual instance", "date": "2026-06-12", "categories": []}
{"id": "2606.11619", "title": "Precision-Aware Illumination-Disentangled Vision Transformer for Spacecraft 6D Pose Estimation", "abstract": "Vision sensors provide a lightweight solution for spacecraft proximity operations, but monocular spacecraft 6D pose estimation remains difficult under illumination variation, specular reflection, shadowing, weak texture, and background interference. These factors make local visual evidence spatially unreliable and can destabilize pose regression. This article proposes a Precision-Aware Illumination-Disentangled Vision Transformer (PAID-ViT) for robust spacecraft pose estimation.The proposed model separates pose-relevant structure tokens from illumination-sensitive appearance tokens, estimates patch reliability before pose aggregation, and uses foreground mask supervision to preserve silhouette cues. A parameter-free geometric recovery module converts normalized crop coordinates, log-depth, and a continuous 6D rotation representation into camera-frame rotation and translation. Experiments on SPEED+ V2, the SPEED+ validation/lightbox/sunlamp evaluation configuration used in this study, suggest that PAID-ViT reduces translation error and improves robustness in the challenging sunlamp domain, while ablation studies support the complementary roles of illumination disentanglement, reliability-aware token aggregation, mask supervision, and training-side regularization.", "date": "2026-06-10", "categories": []}
{"id": "2602.04343", "title": "Finding NeMO: A Geometry-Aware Representation of Template Views for Few-Shot Perception", "abstract": "We present Neural Memory Object (NeMO), a novel object-centric representation that can be used to detect, segment and estimate the 6DoF pose of objects unseen during training using RGB images. Our method consists of an encoder that requires only a few RGB template views depicting an object to generate a sparse object-like point cloud using a learned UDF containing semantic and geometric information. Next, a decoder takes the object encoding together with a query image to generate a variety of dense predictions. Through extensive experiments, we show that our method can be used for few-shot object perception without requiring any camera-specific parameters or retraining on target data. Our proposed concept of outsourcing object information in a NeMO and using a single network for multiple perception tasks enhances interaction with novel objects, improving scalability and efficiency by enabling quick object onboarding without retraining or extensive pre-processing. We report competitive and state-of-the-art results on various datasets and perception tasks of the BOP benchmark, demonstrating the versatility of our approach. https://github.com/DLR-RM/nemo", "date": "2026-02-04", "categories": []}
{"id": "2606.02219", "title": "Symmetry-Aware 9D Pose Estimation with Sim(3)-Consistent Feature and Spherical Inception Convolution", "abstract": "Object pose estimation is a fundamental problem for an agent system to perceive or manipulate objects in images or videos. However, current instance-level methods struggle with generalization to unseen objects. Category-level methods seek to address this, but remain constrained by the complexities of learning in the non-linear Sim(3) space and intra-class variations. To address these challenges, We propose an effective method for category-level object pose estimation with two key innovations: (1) A translation/size estimator, featuring a semantic-guided symmetry-aware module that leverages robust generalization capabilities of a large vision model (LVM) to infer symmetry points, resulting in accurate translation and size without shape priors. This result serves as a precomputed cue for rotation estimation, thereby reducing the difficulty of learning in the non-linear Sim(3) space and laying a robust foundation for tackling the inherently more challenging rotation estimation. (2) A feature fusion module, based on our proposed spherical large-kernel inception convolution, fuses semantic features from the LVM with systematically computed geometric features to extract essential pose features from intra-class variations by modeling long-range dependencies without excessive computational cost. Built on these innovations, we achieve SOTA on benchmarks and real-world scenes, while developing a robust robotic picking system capable of handling diverse objects. Our code will be available at the project page: {\\hypersetup{urlcolor=blue}https://panfei-cheng.github.io/SSH-Pose}.", "date": "2026-06-01", "categories": []}
{"id": "2604.20650", "title": "MAPRPose: Mask-Aware Proposal and Amodal Refinement for Multi-Object 6D Pose Estimation", "abstract": "6D object pose estimation in cluttered scenes remains challenging due to severe occlusion and sensor noise. We propose MAPRPose, a two-stage framework that leverages mask-aware correspondences for pose proposal and amodal-driven Region-of-Interest (ROI) prediction for robust refinement. In the Mask-Aware Pose Proposal (MAPP) stage, we lift 2D correspondences into 3D space to establish reliable keypoint matches and generate geometrically consistent pose hypotheses based on correspondence-level scoring, from which the top-$K$ candidates are selected. In the refinement stage, we introduce a tensorized render-and-compare pipeline integrated with an Amodal Mask Prediction and ROI Re-Alignment (AMPR) module. By reconstructing complete object geometry and dynamically adjusting the ROI, AMPR mitigates localization errors and spatial misalignment under heavy occlusion. Furthermore, our GPU-accelerated RGB-XYZ reprojection enables simultaneous refinement of all $N \\times B$ pose hypotheses in a single forward pass. Evaluated on the BOP benchmark, MAPRPose achieves a state-of-the-art Average Recall (AR) of 76.5%, outperforming FoundationPose by 3.1% AR while delivering a 43x speedup in multi-object inference.", "date": "2026-04-22", "categories": []}
{"id": "2605.31217", "title": "TALON: Token-Aligned Lightweight Adapters for 6-DoF Spacecraft Pose Estimation", "abstract": "Monocular 6-DoF spacecraft pose estimation methods predominantly process individual frames, discarding the temporal information present in an image sequence acquired during spacecraft manoeuvres. Few temporal approaches require full backbone fine-tuning or auxiliary optical flow networks, risking catastrophic forgetting or increasing computational cost, respectively. We propose TALON (Token-Aligned Lightweight adapters for Orbital Navigation): spatiotemporal 3D adapters injected before the self-attention layers of a frozen ViT vision transformer, combined with a patch-token alignment loss that geometrically grounds the adapted features to keypoint structure through a prototype-conditioned KL-divergence objective. Pre-attention placement allows the frozen attention to reason over temporally enriched tokens, achieving stronger performance with a single adapter per block than post-attention alternatives. The alignment loss shapes the intermediate representations so that each keypoint induces a spatially precise activation in the token field, while the framework adds less than 5% parameters to the frozen backbone. On SPADES dataset, TALON reduces the pose error by 50% over the prior state-of-the-art, and on SwissCube dataset it surpasses the prior best by 21.8% in ADD-0.1d accuracy. Zero-shot cross-domain evaluation from sim-to-real on SPARK real data reduces pose error by 4.7x, and ablations characterise the role of adapter depth across in-domain and cross-domain settings.", "date": "2026-05-29", "categories": []}
{"id": "2603.08264", "title": "Event-based Motion & Appearance Fusion for 6D Object Pose Tracking", "abstract": "Object pose tracking is a fundamental and essential task for robotics to perform tasks in the home and industrial settings. The most commonly used sensors to do so are RGB-D cameras, which can hit limitations in highly dynamic environments due to motion blur and frame-rate constraints. Event cameras have remarkable features such as high temporal resolution and low latency, which make them a potentially ideal vision sensors for object pose tracking at high speed. Even so, there are still only few works on 6D pose tracking with event cameras. In this work, we take advantage of the high temporal resolution and propose a method that uses both a propagation step fused with a pose correction strategy. Specifically, we use 6D object velocity obtained from event-based optical flow for pose propagation, after which, a template-based local pose correction module is utilized for pose correction. Our learning-free method has comparable performance to the state-of-the-art algorithms, and in some cases out performs them for fast-moving objects. The results indicate the potential for using event cameras in highly-dynamic scenarios where the use of deep network approaches are limited by low update rates.", "date": "2026-03-09", "categories": []}
{"id": "2605.28270", "title": "Every9D-21M: Large-Scale Real-World 9D Canonicalization of Everyday Objects", "abstract": "Estimating the 9D pose of everyday objects from a single real-world image remains challenging. This is largely due to the lack of large-scale supervision. Most existing datasets either rely heavily on synthetic renderings or provide limited coverage of real-world objects: the largest real-world 9D pose dataset to date contains only 17K annotated objects across 9 categories. We address this gap with Every9D-21M, a dataset of 9D pose annotations for 21.8M real-world images from 109K object- centric videos spanning 700 everyday object categories - two orders of magnitude larger than prior real-world 9D pose benchmarks in both image and category count. To achieve this scale, we leverage object-centric videos by reconstructing object- level point clouds via multi-view geometry and aligning similar instances into a shared canonical coordinate frame. Canonical poses are manually annotated for only a small set of reference objects (fewer than 0.01% of all images) and propagated to the remaining instances via cross-instance alignment. All propagated canonical poses are then verified from multiple viewpoints. We further introduce cross-category orientation rules that induce category-level symmetries, enabling symmetry-aware evaluation. Beyond establishing dedicated training and evaluation splits as a benchmark for 9D pose foundation models, we show that training on Every9D-21M improves performance on ImageNet3D and PASCAL3D+, and generalizes to HANDAL substantially better than training on ImageNet3D. Data and code are available at https://github.com/GenIntel/Every9D.", "date": "2026-05-27", "categories": []}
{"id": "2605.25553", "title": "ComPose: A Unified Completion-Pose Framework for Robust Category-Level Object Pose Estimation", "abstract": "Category-level object pose estimation aims to predict the pose and size of arbitrary objects in specific categories. Existing methods struggle with the inherent incompleteness of observed point clouds, which limits their ability to capture complete object shapes for robust pose reasoning. While point cloud completion offers a promising solution, naively treating it as a separate preprocessing step for partial observations introduces compounding errors and additional computational overhead, ultimately hindering both accuracy and efficiency. To address these challenges, we propose ComPose, a novel unified framework that tightly integrates shape completion to provide complete geometric cues for enhanced pose estimation. At the core of ComPose is a keypoint-based progressive completion module, which recovers full shape representations by progressively predicting a sparse set of keypoints and their surrounding dense point sets, empowering the keypoints to capture holistic object geometries. A geometric relation encoding module further enriches keypoint features with both local and global geometric context. In addition, we introduce a novel geometric relation consistency loss to enforce structural alignment between observed keypoints and their predicted NOCS coordinates, ensuring globally coherent coordinate transformations. Extensive experiments on standard benchmarks demonstrate that our method outperforms state-of-the-art approaches without relying on category-level shape priors.", "date": "2026-05-25", "categories": []}
{"id": "2605.25598", "title": "SurfSurg6D: Geometry Consistent Dense Correspondence for Textureless Surgical Instrument Pose Estimation", "abstract": "Surgical instrument pose estimation provides crucial information for promising applications, including autonomous robotic surgery, skill assessment, and standardization of surgical workflow. However, this task remains highly challenging due to high precision requirements, frequent occlusions, textureless instruments, scarcity of depth information and very limited annotated data. These constraints often lead to unsatisfactory performance when employing general object pose estimation approaches to surgical scenarios. To address these issues, we first construct a new dataset SynSurg6D, to alleviate the data shortage in this task. We further propose SurfSurg6D, a dense-correspondence framework tailored for surgical instrument pose estimation. Experimental results on the SurgRIPE, EndoVis2018 and SurgPose datasets demonstrate that the introduction of our generated dataset SynSurg6D is able to diversify the pose distributions, thus enhancing the performance of existing approaches. Furthermore, SurfSurg6D outperforms existing methods, providing a robust solution for precise and efficient RGB-only pose estimation.", "date": "2026-05-25", "categories": []}
{"id": "2605.23523", "title": "ComPose: When to Trust Hands for Object Pose Tracking", "abstract": "Reconstructing the motion of objects from videos is a key component for embodied AI and robot manipulation. While diverse approaches to object pose tracking have been studied, they rely heavily on strong external priors, such as depth data or 3D templates, and remain highly vulnerable to severe occlusions by hand grasps despite the use of explicit masks. In this work, we present ComPose, a 6DoF object tracking framework designed for hand-aware object pose estimation from RGB video. Rather than treating the hand purely as an occluder, our method harmonizes hand motions as a \\textit{complementary cue} for object tracking. In detail, we recover a variety of object motions over time by combining object and hand cues from foundation models within a unified tracking pipeline. Here, ComPose adaptively selects informative hand joints, combines object- and hand-derived cues for motion estimation, and refines the resulting object motion using visible geometric evidence and a learned correction. We further enforce the temporal consistency over both rotation and translation, yielding stable 3D object trajectories over time without any external smoothing. Extensive experiments show that our method is accurate, efficient, and robust under severe hand occlusion and geometric ambiguity. In addition, the resulting trajectories can also effectively transfer to downstream robot manipulation by enabling robots to reconstruct human actions from online videos.", "date": "2026-05-22", "categories": []}
{"id": "2512.20538", "title": "AlignPose: Generalizable 6D Pose Estimation via Multi-view Feature-metric Alignment", "abstract": "Single-view RGB model-based object pose estimation methods achieve strong generalization but are fundamentally limited by depth ambiguity, clutter, and occlusions. Multi-view pose estimation methods have the potential to solve these issues, but existing works rely on precise single-view pose estimates or lack generalization to unseen objects. We address these challenges via the following three contributions. First, we introduce AlignPose, a 6D object pose estimation method that aggregates information from multiple extrinsically calibrated RGB views and does not require any object-specific training or symmetry annotation. Second, the key component of this approach is a new multi-view feature-metric refinement specifically designed for object pose. It optimizes a single, consistent world-frame object pose by minimizing the feature discrepancy between on-the-fly rendered object features and observed image features across all views simultaneously. Third, we report extensive experiments on six datasets (YCB-V, T-LESS, HouseCat6D, ITODD-MV, IPD, XYZ-IBD) using the BOP benchmark evaluation and show that AlignPose outperforms other published methods, especially on challenging industrial datasets where multiple views are readily available in practice.", "date": "2025-12-23", "categories": []}
{"id": "2605.19649", "title": "CAD-Free Learning of Spacecraft Pose Estimators via NeRF-Based Augmentations", "abstract": "Spacecraft pose estimation networks require tens of thousands of CAD-rendered images to be trained. This reliance on synthetic CAD data (i) limits applicability to targets with reliable geometry prior, excluding uncooperative or poorly documented spacecraft, and (ii) causes poor generalization to real on-orbit conditions due to unrealistic illumination and material appearance. This paper introduces a NeRF-based image augmentation method that enables the learning of spacecraft pose estimators from only a few tens to a few hundreds of images. The method learns a Neural Radiance Field of the target and generates a large, diverse dataset through geometrically-consistent viewpoint and appearance augmentation. This augmented dataset enables the training of accurate target-specific pose estimators without requiring a CAD model or large synthetic datasets. Experiments show that our approach supports the training of accurate pose estimators from only 25 to 400 realistic images, even under severe illumination variations. When applied on large CAD-based synthetic datasets, the NeRF-based augmentation also enhances out-of-domain generalization, yielding improved robustness to real on-orbit conditions.", "date": "2026-05-19", "categories": []}
{"id": "2605.15728", "title": "DecomPose: Disentangling Cross-Category Optimization Contention for Category-Level 6D Object Pose Estimation", "abstract": "Category-level 6D object pose estimation is typically formulated as a multi-category joint learning problem with fully shared model parameters. However, pronounced geometric heterogeneity across categories entangles incompatible optimization signals in shared modules, resulting in gradient conflicts and negative transfer during training. To address this challenge, we first introduce gradient-based diagnostics to quantify module-level cross-category contention. Building on results of diagnostics, we propose DecomPose, a difficulty-aware decomposition framework that mitigates optimization contention via: (1) difficulty-aware gradient decoupling, which groups categories using a data-driven difficulty proxy and routes each instance to a group-specific correspondence branch to isolate incompatible updates; and (2) stability-driven asymmetric branching, which assigns higher-capacity branches to structurally simple categories as stable optimization anchors while constraining complex categories with lightweight branches to suppress noisy updates and alleviate negative transfer. Extensive experiments on REAL275, CAMERA25, and HouseCat6D demonstrate that DecomPose effectively reduces cross-category optimization contention and delivers superior pose estimation performance across multiple benchmarks.", "date": "2026-05-15", "categories": []}
{"id": "2511.09771", "title": "STORM: Segment, Track, and Object Re-Localization from a Single Image", "abstract": "Accurate 6D pose estimation and tracking are core capabilities for physical AI systems, yet real-world deployment remains brittle and labor-intensive. Many pipelines rely on CAD models, manual masking, or per-object adaptation, and still fail under occlusion or fast motion without a principled way to recognize failure. We propose STORM, a unified framework for reference-conditioned 6D tracking that can operate from a single reference image, with minimal manual input and improved robustness. STORM combines: (i) Hierarchical Spatial Fusion Attention (HSFA), a task-driven reference-query fusion architecture that supports both single-reference and multi-reference conditioning and can optionally use vision-language semantic conditioning to resolve instance ambiguities; and (ii) a BCE-trained tracking verifier whose continuous compatibility logit is used as an energy-like score to detect drift and trigger automatic re-initialization. Experiments on LM-O and YCB-Video show that STORM improves annotation-free pose tracking accuracy over strong baselines and recovers reliably from severe occlusions and rapid viewpoint changes with minimal overhead.", "date": "2025-11-12", "categories": []}
{"id": "2604.06720", "title": "Exploring 6D Object Pose Estimation with Deformation", "abstract": "We present DeSOPE, a large-scale dataset for 6DoF deformed objects. Most 6D object pose methods assume rigid or articulated objects, an assumption that fails in practice as objects deviate from their canonical shapes due to wear, impact, or deformation. To model this, we introduce the DeSOPE dataset, which features high-fidelity 3D scans of 26 common object categories, each captured in one canonical state and three deformed configurations, with accurate 3D registration to the canonical mesh. Additionally, it features an RGB-D dataset with 133K frames across diverse scenarios and 665K pose annotations produced via a semi-automatic pipeline. We begin by annotating 2D masks for each instance, then compute initial poses using an object pose method, refine them through an object-level SLAM system, and finally perform manual verification to produce the final annotations. We evaluate several object pose methods and find that performance drops sharply with increasing deformation, suggesting that robust handling of such deformations is critical for practical applications.", "date": "2026-04-08", "categories": []}
{"id": "2605.08592", "title": "Cross-Modal RGB-D Fusion Transformer for 6D Pose Estimation of Non-Cooperative Spacecraft with Stereo-Derived Depth", "abstract": "On-orbit servicing and active debris removal involving non-cooperative spacecraft require reliable pose estimation to supply accurate position and orientation data for autonomous visual navigation. Learning-based monocular methods have seen widespread adoption in spacecraft pose estimation, yet they suffer from an intrinsic depth ambiguity problem and tend to fail under the harsh illumination conditions routinely encountered in orbit. Active depth sensors could in principle address the geometric ambiguity, but their power and mass requirements make them poorly suited to most spacecraft platforms. This work addresses these issues through a passive stereo vision framework for six-degree-of-freedom (6-DOF) pose estimation of non-cooperative spacecraft. A binocular stereo matching network called TSCA-Stereo is developed to cope with weak-texture surfaces, specular highlights, and severe lighting variations typical of space imagery. A cross-modal fusion Transformer is introduced to combine RGB appearance information with stereo depth features in an adaptive manner, supporting reliable pose recovery. A synthetic binocular multimodal dataset is also built for the experiments, covering stereo disparity maps and 6-DOF pose annotations across a range of illumination scenarios, attitude configurations, and noise levels. Experimental results show that TSCA-Stereo outperforms the baseline across every evaluated metric on this space-specific dataset. The full pose estimation pipeline achieves a mean translation error of 0.0419 m and a mean orientation error of 0.8632{\\deg} under varied imaging conditions, confirming that the passive stereo approach is both effective and resilient when operating under the demanding visual conditions of the space environment.", "date": "2026-05-09", "categories": []}
{"id": "2605.07023", "title": "OneViewAll: Semantic Prior Guided One-View 6D Pose Estimation for Novel Objects", "abstract": "In many practical 6D object pose estimation scenarios, we often have access to only a single real-world RGB-D reference view per object, typically without CAD models. Existing methods largely rely on explicit 3D models or multi-view data, which limits their scalability. To address this challenging single-reference model-free setting, we propose \\textbf{OneViewAll}, a semantic-prior-guided framework that performs pose estimation via a novel Project-and-Compare paradigm. Instead of relying on computationally expensive CAD-based rendering, our method directly aligns reference and query observations within a projection-equivariant space. OneViewAll progressively integrates hierarchical semantic priors across three levels: (1) \\textit{category- and scene-level} priors for efficient hypothesis initialization; (2) \\textit{object-level symmetry} priors for geometry completion via mirror fusion; and (3) \\textit{patch-level} priors for discriminative refinement. Extensive experiments demonstrate that OneViewAll achieves \\textbf{92.5\\%} ADD-0.1 accuracy on the LINEMOD dataset using only one real reference view -- significantly outperforming the CVPR 2025 baseline One2Any (52.6\\%). It also yields consistent improvements on YCB-V, Real275, and Toyota-Light while maintaining low inference latency. Our results underscore the efficacy of symmetry-aware projection in handling symmetric, texture-less, and occluded objects.", "date": "2026-05-07", "categories": []}
{"id": "2605.08059", "title": "6D Pose Estimation via Keypoint Heatmap Regression with RGB-D Residual Neural Networks", "abstract": "In this paper, we propose a modular framework for 6D pose estimation based on keypoint heatmap regression. Our approach combines YOLOv10m for object detection with a ResNet18-based network that predicts 2D heatmaps from RGB images. Keypoints extracted from these heatmaps are used to estimate the 6D object pose via the PnP RANSAC algorithm. We compare different keypoint selection strategies to assess their impact on pose accuracy. Additionally, we extend the baseline by incorporating depth data using a cross-fusion architecture, which enables interaction between RGB and depth features at multiple stages. We further explore general training improvements, such as experimenting with activation functions and learning rate scheduling strategies to improve model performance. Our best RGB-only model achieved a mean ADD-based accuracy of 84.50%, while the RGB-D fusion model reached 92.41% on the LINEMOD dataset. The code is available at https://github.com/ameermasood/HeatNet.", "date": "2026-05-08", "categories": []}
{"id": "2605.02708", "title": "Temporally Consistent Object 6D Pose Estimation for Robot Control", "abstract": "Single-view RGB object pose estimators have reached a level of precision and efficiency that makes them good candidates for vision-based robot control. However, off-the-shelf methods lack temporal consistency and robustness that are mandatory for a stable feedback control. In this work, we develop a factor graph approach to enforce temporal consistency of the object pose estimates. In particular, the proposed approach: (i) incorporates object motion models, (ii) explicitly estimates the object pose measurement uncertainty, and (iii) integrates the above two components in an online optimization-based estimator. We demonstrate that with appropriate outlier rejection and smoothing using the proposed factor graph approach, we can significantly improve the results on standardized pose estimation benchmarks. We experimentally validate the stability of the proposed approach for a feedback-based robot control task in which the object is tracked by the camera attached to a torque controlled manipulator.", "date": "2026-05-04", "categories": []}
{"id": "2512.06565", "title": "GNC-Pose: Geometry-Aware GNC-PnP for Accurate 6D Pose Estimation", "abstract": "We present GNC-Pose, a fully learning-free monocular 6D object pose estimation pipeline for textured objects that combines rendering-based initialization, geometry-aware correspondence weighting, and robust GNC optimization. Starting from coarse 2D-3D correspondences obtained through feature matching and rendering-based alignment, our method builds upon the Graduated Non-Convexity (GNC) principle and introduces a geometry-aware, cluster-based weighting mechanism that assigns robust per point confidence based on the 3D structural consistency of the model. This geometric prior and weighting strategy significantly stabilizes the optimization under severe outlier contamination. A final LM refinement further improve accuracy. We tested GNC-Pose on The YCB Object and Model Set, despite requiring no learned features, training data, or category-specific priors, GNC-Pose achieves competitive accuracy compared with both learning-based and learning-free methods, and offers a simple, robust, and practical solution for learning-free 6D pose estimation.", "date": "2025-12-06", "categories": []}
{"id": "2604.23387", "title": "Keypoint-based Dynamic Object 6-DoF Pose Tracking via Event Camera", "abstract": "Accurate 6-DoF pose estimation of objects is critical for robots to perform precise manipulation tasks. However, for dynamic object pose estimation, conventional camera-based approaches face several major challenges, such as motion blur, sensor noise, and low-light limitation. To address these issues, we employ event cameras, whose high dynamic range and low latency offer a promising solution. Furthermore, we propose a keypoint-based detection and tracking approach for dynamic object pose estimation. Firstly, a keypoint detection network is constructed to extract keypoints from the time surface generated by the event stream. Subsequently, the polarity and spatial coordinates of the events are leveraged, and the event density in the vicinity of each keypoint is utilized to achieve continuous keypoint tracking. Finally, a hash mapping is established between the 2D keypoints and the 3D model keypoints, and the EPnP algorithm is employed to estimate the 6-DoF pose. Experimental results demonstrate that, whether in simulated or real event environments, the proposed method outperforms the event-based state-of-the-art methods in terms of both accuracy and robustness.", "date": "2026-04-25", "categories": []}
{"id": "2604.16954", "title": "TSM-Pose: Topology-Aware Learning with Semantic Mamba for Category-Level Object Pose Estimation", "abstract": "Category-level object pose estimation is fundamental for embodied intelligence, yet achieving robust generalization to unseen instances remains challenging. However, existing methods mainly rely on simple feature extraction and aggregation, which struggle to capture category-shared topological structures and conduct semantic keypoint modeling, limiting their generalization. To address these, we propose a \\textbf{T}opology-Aware Learning with \\textbf{S}emantic \\textbf{M}amba for Category-Level \\textbf{P}ose Estimation framework (TSM-Pose). Specifically, we introduce a Topology Extractor to capture the global topological representation of the point cloud, which is integrated into local geometry features and enables robust category-level structural representation. Simultaneously, we propose a Mamba-based Global Semantic Aggregator that injects semantics priors into keypoints to enhance their expressiveness and leverages multiple TwinMamba blocks to model long-range dependencies for more effective global feature aggregation. Extensive experiments on three benchmark datasets (REAL275, CAMERA25, and HouseCat6D) demonstrate that TSM-Pose outperforms existing state-of-the-art methods.", "date": "2026-04-18", "categories": []}
{"id": "2604.18208", "title": "Towards Symmetry-sensitive Pose Estimation: A Rotation Representation for Symmetric Object Classes", "abstract": "Symmetric objects are common in daily life and industry, yet their inherent orientation ambiguities that impede the training of deep learning networks for pose estimation are rarely discussed in the literature. To cope with these ambiguities, existing solutions typically require the design of specific loss functions and network architectures or resort to symmetry-invariant evaluation metrics. In contrast, we focus on the numeric representation of the rotation itself, modifying trigonometric identities with the degrees of symmetry derived from the objects' shapes. We use our representation, SARR, to obtain canonic (symmetry-resolved) poses for the symmetric objects in two popular 6D pose estimation datasets, T-LESS and ITODD, where SARR is unique and continuous w.r.t. the visual appearance. This allows us to use a standard CNN for 3D orientation estimation whose performance is evaluated with the symmetry-sensitive cosine distance $\\text{AR}_{\\text{C}}$. Our networks outperform the state of the art using $\\text{AR}_{\\text{C}}$ and achieve satisfactory performance when using conventional symmetry-invariant measures. Our method does not require any 3D models but only depth, or, as part of an additional experiment, texture-less RGB/grayscale images as input. We also show that networks trained on SARR outperform the same networks trained on rotation matrices, Euler angles, quaternions, standard trigonometrics or the recently popular 6d representation -- even in inference scenarios where no prior knowledge of the objects' symmetry properties is available. Code and a visualization toolkit are available at https://github.com/akriegler/SARR .", "date": "2026-04-20", "categories": []}
{"id": "2503.05578", "title": "Scalable Unseen Objects 6-DoF Absolute Pose Estimation with Robotic Integration", "abstract": "Pose estimation-guided unseen object 6-DoF robotic manipulation is a key task in robotics. However, the scalability of current pose estimation methods to unseen objects remains a fundamental challenge, as they generally rely on CAD models or dense reference views of unseen objects, which are difficult to acquire, ultimately limit their scalability. In this paper, we introduce a novel task setup, referred to as SinRef-6D, which addresses 6-DoF absolute pose estimation for unseen objects using only a single pose-labeled reference RGB-D image captured during robotic manipulation. This setup is more scalable yet technically nontrivial due to large pose discrepancies and the limited geometric and spatial information contained in a single view. To address these issues, our key idea is to iteratively establish point-wise alignment in a common coordinate system with state space models (SSMs) as backbones. Specifically, to handle large pose discrepancies, we introduce an iterative object-space point-wise alignment strategy. Then, Point and RGB SSMs are proposed to capture long-range spatial dependencies from a single view, offering superior spatial modeling capability with linear complexity. Once pre-trained on synthetic data, SinRef-6D can estimate the 6-DoF absolute pose of an unseen object using only a single reference view. With the estimated pose, we further develop a hardware-software robotic system and integrate the proposed SinRef-6D into it in real-world settings. Extensive experiments on six benchmarks and in diverse real-world scenarios demonstrate that our SinRef-6D offers superior scalability. Additional robotic grasping experiments further validate the effectiveness of the developed robotic system. The code and robotic demos are available at https://paperreview99.github.io/SinRef-6DoF-Robotic.", "date": "2025-03-07", "categories": []}
{"id": "2511.07717", "title": "RoboTAG: End-to-end Robot Configuration Estimation via Topological Alignment Graph", "abstract": "Estimating robot pose from a monocular RGB image is a challenge in robotics and computer vision. Existing methods typically build networks on top of 2D visual backbones and depend heavily on labeled data for training, which is often scarce in real-world scenarios, causing a sim-to-real gap. Moreover, these approaches reduce the 3D-based problem to 2D domain, neglecting the 3D priors. To address these, we propose Robot Topological Alignment Graph (RoboTAG), which incorporates a 3D branch to inject 3D priors while enabling co-evolution of the 2D and 3D representations, alleviating the reliance on labels. Specifically, the RoboTAG consists of a 3D branch and a 2D branch, where nodes represent the states of the camera and robot system, and edges capture the dependencies between these variables or denote alignments between them. Closed loops are then defined in the graph, on which a consistency supervision across branches can be applied. Experimental results demonstrate that our method is effective across robot types, suggesting new possibilities of alleviating the data bottleneck in robotics.", "date": "2025-11-11", "categories": []}
{"id": "2512.09056", "title": "ConceptPose: Training-Free Zero-Shot Object Pose Estimation using Concept Vectors", "abstract": "Object pose estimation is a fundamental task in computer vision and robotics, yet most methods require extensive, dataset-specific training. Concurrently, large-scale vision language models show remarkable zero-shot capabilities. In this work, we bridge these two worlds by introducing ConceptPose, a framework for object pose estimation that is both training-free and model-free. ConceptPose leverages a vision-language-model (VLM) to create open-vocabulary 3D concept maps, where each point is tagged with a concept vector derived from saliency maps. By establishing robust 3D-3D correspondences across concept maps, our approach allows precise estimation of 6DoF relative pose. Without any object or dataset-specific training, our approach achieves state-of-the-art results on common zero shot relative pose estimation benchmarks, outperforming the strongest baseline by a relative 62\\% in average ADD(-S) score, including methods that utilize extensive dataset-specific training.", "date": "2025-12-09", "categories": []}
{"id": "2604.10415", "title": "Point2Pose: Occlusion-Recovering 6D Pose Tracking and 3D Reconstruction for Multiple Unknown Objects Via 2D Point Trackers", "abstract": "We present Point2Pose, a model-free method for causal 6D pose tracking of multiple rigid objects from monocular RGB-D video. Initialized only from sparse image points on the objects to be tracked, our approach tracks multiple unseen objects without requiring object CAD models or category priors. Point2Pose leverages a 2D point tracker to obtain long-range correspondences, enabling instant recovery after complete occlusion. Simultaneously, the system incrementally reconstructs an online Truncated Signed Distance Function (TSDF) representation of the tracked targets. Alongside the method, we introduce a new multi-object tracking dataset comprising both simulation and real-world sequences, with motion-capture ground truth for evaluation. Experiments show that Point2Pose achieves performance comparable to the state-of-the-art methods on a severe-occlusion benchmark, while additionally supporting multi-object tracking and recovery from complete occlusion, capabilities that are not supported by previous model-free tracking approaches.", "date": "2026-04-12", "categories": []}
{"id": "2604.09199", "title": "Globally Optimal Pose from Orthographic Silhouettes", "abstract": "We solve the problem of determining the pose of known shapes in $\\mathbb{R}^3$ from their unoccluded silhouettes. The pose is determined up to global optimality using a simple yet under-explored property of the area-of-silhouette: its continuity w.r.t trajectories in the rotation space. The proposed method utilises pre-computed silhouette-signatures, modelled as a response surface of the area-of-silhouettes. Querying this silhouette-signature response surface for pose estimation leads to a strong branching of the rotation search space, making resolution-guided candidate search feasible. Additionally, we utilise the aspect ratio of 2D ellipses fitted to projected silhouettes as an auxiliary global shape signature to accelerate the pose search. This combined strategy forms the first method to efficiently estimate globally optimal pose from just the silhouettes, without being guided by correspondences, for any shape, irrespective of its convexity and genus. We validate our method on synthetic and real examples, demonstrating significantly improved accuracy against comparable approaches.\n  Code, data, and supplementary in: https://agnivsen.github.io/pose-from-silhouette/", "date": "2026-04-10", "categories": []}
{"id": "2603.28045", "title": "Event6D: Event-based Novel Object 6D Pose Tracking", "abstract": "Event cameras provide microsecond latency, making them suitable for 6D object pose tracking in fast, dynamic scenes where conventional RGB and depth pipelines suffer from motion blur and large pixel displacements. We introduce EventTrack6D, an event-depth tracking framework that generalizes to novel objects without object-specific training by reconstructing both intensity and depth at arbitrary timestamps between depth frames. Conditioned on the most recent depth measurement, our dual reconstruction recovers dense photometric and geometric cues from sparse event streams. Our EventTrack6D operates at over 120 FPS and maintains temporal consistency under rapid motion. To support training and evaluation, we introduce a comprehensive benchmark suite: a large-scale synthetic dataset for training and two complementary evaluation sets, including real and simulated event datasets. Trained exclusively on synthetic data, EventTrack6D generalizes effectively to real-world scenarios without fine-tuning, maintaining accurate tracking across diverse objects and motion patterns. Our method and datasets validate the effectiveness of event cameras for event-based 6D pose tracking of novel objects. Code and datasets are publicly available at https://chohoonhee.github.io/Event6D.", "date": "2026-03-30", "categories": []}
{"id": "2603.27429", "title": "Mind the Shape Gap: A Benchmark and Baseline for Deformation-Aware 6D Pose Estimation of Agricultural Produce", "abstract": "Accurate 6D pose estimation for robotic harvesting is fundamentally hindered by the biological deformability and high intra-class shape variability of agricultural produce. Instance-level methods fail in this setting, as obtaining exact 3D models for every unique piece of produce is practically infeasible, while category-level approaches that rely on a fixed template suffer significant accuracy degradation when the prior deviates from the true instance geometry. To bridge such lack of robustness to deformation, we introduce PEAR (Pose and dEformation of Agricultural pRoduce), the first benchmark providing joint 6D pose and per-instance 3D deformation ground truth across 8 produce categories, acquired via a robotic manipulator for high annotation accuracy. Using PEAR, we show that state-of-the-art methods suffer up to 6x performance degradation when faced with the inherent geometric deviations of real-world produce. Motivated by this finding, we propose SEED (Simultaneous Estimation of posE and Deformation), a unified RGB-only framework that jointly predicts 6D pose and explicit lattice deformations from a single image across multiple produce categories. Trained entirely on synthetic data with generative texture augmentation applied at the UV level, SEED outperforms MegaPose on 6 out of 8 categories under identical RGB-only conditions, demonstrating that explicit shape modeling is a critical step toward reliable pose estimation in agricultural robotics.", "date": "2026-03-28", "categories": []}
{"id": "2603.27533", "title": "Demo-Pose: Depth-Monocular Modality Fusion For Object Pose Estimation", "abstract": "Object pose estimation is a fundamental task in 3D vision with applications in robotics, AR/VR, and scene understanding. We address the challenge of category-level 9-DoF pose estimation (6D pose + 3Dsize) from RGB-D input, without relying on CAD models during inference. Existing depth-only methods achieve strong results but ignore semantic cues from RGB, while many RGB-D fusion models underperform due to suboptimal cross-modal fusion that fails to align semantic RGB cues with 3D geometric representations. We propose DeMo-Pose, a hybrid architecture that fuses monocular semantic features with depth-based graph convolutional representations via a novel multimodal fusion strategy. To further improve geometric reasoning, we introduce a novel Mesh-Point Loss (MPL) that leverages mesh structure during training without adding inference overhead. Our approach achieves real-time inference and significantly improves over state-of-the-art methods across object categories, outperforming the strong GPV-Pose baseline by 3.2\\% on 3D IoU and 11.1\\% on pose accuracy on the REAL275 benchmark. The results highlight the effectiveness of depth-RGB fusion and geometry-aware learning, enabling robust category-level 3D pose estimation for real-world applications.", "date": "2026-03-29", "categories": []}
{"id": "2603.19961", "title": "Cov2Pose: Leveraging Spatial Covariance for Direct Manifold-aware 6-DoF Object Pose Estimation", "abstract": "In this paper, we address the problem of 6-DoF object pose estimation from a single RGB image. Indirect methods that typically predict intermediate 2D keypoints, followed by a Perspective-n-Point solver, have shown great performance. Direct approaches, which regress the pose in an end-to-end manner, are usually computationally more efficient but less accurate. However, direct pose regression heads rely on globally pooled features, ignoring spatial second-order statistics despite their informativeness in pose prediction. They also predict, in most cases, discontinuous pose representations that lack robustness. Herein, we therefore propose a covariance-pooled representation that encodes convolutional feature distributions as a symmetric positive definite (SPD) matrix. Moreover, we propose a novel pose encoding in the form of an SPD matrix via its Cholesky decomposition. Pose is then regressed in an end-to-end manner with a manifold-aware network head, taking into account the Riemannian geometry of SPD matrices. Experiments and ablations consistently demonstrate the relevance of second-order pooling and continuous representations for direct pose regression, including under partial occlusion.", "date": "2026-03-20", "categories": []}
{"id": "2603.25135", "title": "EgoXtreme: A Dataset for Robust Object Pose Estimation in Egocentric Views under Extreme Conditions", "abstract": "Smart glass is emerging as an useful device since it provides plenty of insights under hands-busy, eyes-on-task situations. To understand the context of the wearer, 6D object pose estimation in egocentric view is becoming essential. However, existing 6D object pose estimation benchmarks fail to capture the challenges of real-world egocentric applications, which are often dominated by severe motion blur, dynamic illumination, and visual obstructions. This discrepancy creates a significant gap between controlled lab data and chaotic real-world application. To bridge this gap, we introduce EgoXtreme, a new large-scale 6D pose estimation dataset captured entirely from an egocentric perspective. EgoXtreme features three challenging scenarios - industrial maintenance, sports, and emergency rescue - designed to introduce severe perceptual ambiguities through extreme lighting, heavy motion blur, and smoke. Evaluations of state-of-the-art generalizable pose estimators on EgoXtreme indicate that their generalization fails to hold in extreme conditions, especially under low light. We further demonstrate that simply applying image restoration (e.g., deblurring) offers no positive improvement for extreme conditions. While performance gain has appeared in tracking-based approach, implying using temporal information in fast-motion scenarios is meaningful. We conclude that EgoXtreme is an essential resource for developing and evaluating the next generation of pose estimation models robust enough for real-world egocentric vision. The dataset and code are available at https://taegyoun88.github.io/EgoXtreme/", "date": "2026-03-26", "categories": []}
{"id": "2603.25228", "title": "Training-free Detection and 6D Pose Estimation of Unseen Surgical Instruments", "abstract": "Purpose: Accurate detection and 6D pose estimation of surgical instruments are crucial for many computer-assisted interventions. However, supervised methods lack flexibility for new or unseen tools and require extensive annotated data. This work introduces a training-free pipeline for accurate multi-view 6D pose estimation of unseen surgical instruments, which only requires a textured CAD model as prior knowledge. Methods: Our pipeline consists of two main stages. First, for detection, we generate object mask proposals in each view and score their similarity to rendered templates using a pre-trained feature extractor. Detections are matched across views, triangulated into 3D instance candidates, and filtered using multi-view geometric consistency. Second, for pose estimation, a set of pose hypotheses is iteratively refined and scored using feature-metric scores with cross-view attention. The best hypothesis undergoes a final refinement using a novel multi-view, occlusion-aware contour registration, which minimizes reprojection errors of unoccluded contour points. Results: The proposed method was rigorously evaluated on real-world surgical data from the MVPSP dataset. The method achieves millimeter-accurate pose estimates that are on par with supervised methods under controlled conditions, while maintaining full generalization to unseen instruments. These results demonstrate the feasibility of training-free, marker-less detection and tracking in surgical scenes, and highlight the unique challenges in surgical environments. Conclusion: We present a novel and flexible pipeline that effectively combines state-of-the-art foundational models, multi-view geometry, and contour-based refinement for high-accuracy 6D pose estimation of surgical instruments without task-specific training. This approach enables robust instrument tracking and scene understanding in dynamic clinical environments.", "date": "2026-03-26", "categories": []}
{"id": "2603.23370", "title": "Object Pose Transformer: Unifying Unseen Object Pose Estimation", "abstract": "Learning model-free object pose estimation for unseen instances remains a fundamental challenge in 3D vision. Existing methods typically fall into two disjoint paradigms: category-level approaches predict absolute poses in a canonical space but rely on predefined taxonomies, while relative pose methods estimate cross-view transformations but cannot recover single-view absolute pose. In this work, we propose Object Pose Transformer (\\ours{}), a unified feed-forward framework that bridges these paradigms through task factorization within a single model. \\ours{} jointly predicts depth, point maps, camera parameters, and normalized object coordinates (NOCS) from RGB inputs, enabling both category-level absolute SA(3) pose and unseen-object relative SE(3) pose. Our approach leverages contrastive object-centric latent embeddings for canonicalization without requiring semantic labels at inference time, and uses point maps as a camera-space representation to enable multi-view relative geometric reasoning. Through cross-frame feature interaction and shared object embeddings, our model leverages relative geometric consistency across views to improve absolute pose estimation, reducing ambiguity in single-view predictions. Furthermore, \\ours{} is camera-agnostic, learning camera intrinsics on-the-fly and supporting optional depth input for metric-scale recovery, while remaining fully functional in RGB-only settings. Extensive experiments on diverse benchmarks (NOCS, HouseCat6D, Omni6DPose, Toyota-Light) demonstrate state-of-the-art performance in both absolute and relative pose estimation tasks within a single unified architecture.", "date": "2026-03-24", "categories": []}
{"id": "2508.03243", "title": "MVTOP: Multi-View Transformer-based Object Pose-Estimation", "abstract": "We present MVTOP, a novel transformer-based method for multi-view rigid object pose estimation. Through an early fusion of the view-specific features, our method can resolve pose ambiguities that would be impossible to solve with a single view or with a post-processing of single-view poses. MVTOP models the multi-view geometry via lines of sight that emanate from the respective camera centers. While the method assumes the camera interior and relative orientations are known for a particular scene, they can vary for each inference. This makes the method versatile. The use of the lines of sight enables MVTOP to correctly predict the correct pose with the merged multi-view information. To show the model's capabilities, we provide a synthetic data set that can only be solved with such holistic multi-view approaches since the poses in the dataset cannot be solved with just one view. Our method outperforms single-view and all existing multi-view approaches on our dataset and achieves competitive results on the YCB-V dataset. To the best of our knowledge, no holistic multi-view method exists that can resolve such pose ambiguities reliably. Our model is end-to-end trainable and does not require any additional data, e.g., depth.", "date": "2025-08-05", "categories": []}
{"id": "2507.05751", "title": "SenseShift6D: Multimodal RGB-D Benchmarking for Robust 6D Pose Estimation across Environment and Sensor Variations", "abstract": "Recent advances on 6D object pose estimation have achieved high performance on representative benchmarks such as LM-O, YCB-V, and T-Less. However, these datasets were captured under fixed illumination and camera settings, leaving the impact of real-world variations in illumination, exposure, gain or depth-sensor mode largely unexplored. To bridge this gap, we introduce SenseShift6D, the first RGB-D dataset that physically sweeps 13 RGB exposures, 9 RGB gains, auto-exposure, 4 depth-capture modes, and 5 illumination levels. For six common household objects, we acquire 198.8k RGB and 20.0k depth images (i.e., 795.4k RGB-D scenes), providing 1,380 unique sensor-lighting permutations per object pose. Experiments with state-of-the-art pretrained, generalizable pose estimators reveal substantial performance variation across lighting and sensor settings, despite their large-scale pretraining. Strikingly, even instance-level estimators-trained and tested on identical objects and backgrounds-exhibit pronounced sensitivity to environmental and sensor shifts. These findings establish sensor- and environment-aware robustness as an underexplored yet essential dimension for real-world deployment, and motivate SenseShift6D as a necessary benchmark for the community. Finally, to illustrate the opportunity enabled by this benchmark, we evaluate test-time multimodal sensor selection without retraining. An idealized (oracle) controller yields remarkable gains of up to +16.7 pp for generalizable models, whereas a practical consistency-based proxy improves performance only marginally, highlighting substantial headroom and the need for future research on reliable sensor-aware adaptation.", "date": "2025-07-08", "categories": []}
{"id": "2312.11894", "title": "3D-LFM: Lifting Foundation Model", "abstract": "The lifting of 3D structure and camera from 2D landmarks is at the cornerstone of the entire discipline of computer vision. Traditional methods have been confined to specific rigid objects, such as those in Perspective-n-Point (PnP) problems, but deep learning has expanded our capability to reconstruct a wide range of object classes (e.g. C3DPO and PAUL) with resilience to noise, occlusions, and perspective distortions. All these techniques, however, have been limited by the fundamental need to establish correspondences across the 3D training data -- significantly limiting their utility to applications where one has an abundance of \"in-correspondence\" 3D data. Our approach harnesses the inherent permutation equivariance of transformers to manage varying number of points per 3D data instance, withstands occlusions, and generalizes to unseen categories. We demonstrate state of the art performance across 2D-3D lifting task benchmarks. Since our approach can be trained across such a broad class of structures we refer to it simply as a 3D Lifting Foundation Model (3D-LFM) -- the first of its kind.", "date": "2023-12-19", "categories": []}
{"id": "2508.14965", "title": "You Only Pose Once: A Minimalist's Detection Transformer for Monocular RGB Category-level 9D Multi-Object Pose Estimation", "abstract": "Accurately recovering the full 9-DoF pose of unseen instances within specific categories from a single RGB image remains a core challenge for robotics and automation. Most existing solutions still rely on pseudo-depth, CAD models, or multi-stage cascades that separate 2D detection from pose estimation. Motivated by the need for a simpler, RGB-only alternative that learns directly at the category level, we revisit a longstanding question: Can object detection and 9-DoF pose estimation be unified with high performance, without any additional data? We show that they can with our method, YOPO, a single-stage, query-based framework that treats category-level 9-DoF estimation as a natural extension of 2D detection. YOPO augments a transformer detector with a lightweight pose head, a bounding-box-conditioned translation module, and a 6D-aware Hungarian matching cost. The model is trained end-to-end only with RGB images and category-level pose labels. Despite its minimalist design, YOPO sets a new state of the art on three benchmarks. On the REAL275 dataset, it achieves 79.6% $\\rm{IoU}_{50}$ and 54.1% under the $10^\\circ$$10{\\rm{cm}}$ metric, surpassing prior RGB-only methods and closing much of the gap to RGB-D systems. The code, models, and additional qualitative results can be found on https://mikigom.github.io/YOPO-project-page.", "date": "2025-08-20", "categories": []}
{"id": "2603.03879", "title": "Yolo-Key-6D: Single Stage Monocular 6D Pose Estimation with Keypoint Enhancements", "abstract": "Estimating the 6D pose of objects from a single RGB image is a critical task for robotics and extended reality applications. However, state-of-the-art multi stage methods often suffer from high latency, making them unsuitable for real time use. In this paper, we present Yolo-Key-6D, a novel single stage, end-to-end framework for monocular 6D pose estimation designed for both speed and accuracy. Our approach enhances a YOLO based architecture by integrating an auxiliary head that regresses the 2D projections of an object's 3D bounding box corners. This keypoint detection task significantly improves the network's understanding of 3D geometry. For stable end-to-end training, we directly regress rotation using a continuous 9D representation projected to SO(3) via singular value decomposition. On the LINEMOD and LINEMOD-Occluded benchmarks, YOLO-Key-6D achieves competitive accuracy scores of 96.24% and 69.41%, respectively, with the ADD(-S) 0.1d metric, while proving itself to operate in real time. Our results demonstrate that a carefully designed single stage method can provide a practical and effective balance of performance and efficiency for real world deployment.", "date": "2026-03-04", "categories": []}
{"id": "2603.00493", "title": "COG: Confidence-aware Optimal Geometric Correspondence for Unsupervised Single-reference Novel Object Pose Estimation", "abstract": "Estimating the 6DoF pose of a novel object with a single reference view is challenging due to occlusions, view-point changes, and outliers. A core difficulty lies in finding robust cross-view correspondences, as existing methods often rely on discrete one-to-one matching that is non-differentiable and tends to collapse onto sparse key-points. We propose Confidence-aware Optimal Geometric Correspondence (COG), an unsupervised framework that formulates correspondence estimation as a confidence-aware optimal transport problem. COG produces balanced soft correspondences by predicting point-wise confidences and injecting them as optimal transport marginals, suppressing non-overlapping regions. Semantic priors from vision foundation models further regularize the correspondences, leading to stable pose estimation. This design integrates confidence into the correspondence finding and pose estimation pipeline, enabling unsupervised learning. Experiments show unsupervised COG achieves comparable performance to supervised methods, and supervised COG outperforms them.", "date": "2026-02-28", "categories": []}
{"id": "2602.19565", "title": "DICArt: Advancing Category-level Articulated Object Pose Estimation in Discrete State-Spaces", "abstract": "Articulated object pose estimation is a core task in embodied AI. Existing methods typically regress poses in a continuous space, but often struggle with 1) navigating a large, complex search space and 2) failing to incorporate intrinsic kinematic constraints. In this work, we introduce DICArt (DIsCrete Diffusion for Articulation Pose Estimation), a novel framework that formulates pose estimation as a conditional discrete diffusion process. Instead of operating in a continuous domain, DICArt progressively denoises a noisy pose representation through a learned reverse diffusion procedure to recover the GT pose. To improve modeling fidelity, we propose a flexible flow decider that dynamically determines whether each token should be denoised or reset, effectively balancing the real and noise distributions during diffusion. Additionally, we incorporate a hierarchical kinematic coupling strategy, estimating the pose of each rigid part hierarchically to respect the object's kinematic structure. We validate DICArt on both synthetic and real-world datasets. Experimental results demonstrate its superior performance and robustness. By integrating discrete generative modeling with structural priors, DICArt offers a new paradigm for reliable category-level 6D pose estimation in complex environments.", "date": "2026-02-23", "categories": []}
{"id": "2602.19719", "title": "Generative 6D Pose Estimation via Conditional Flow Matching", "abstract": "Existing methods for instance-level 6D pose estimation typically rely on neural networks that either directly regress the pose in $\\mathrm{SE}(3)$ or estimate it indirectly via local feature matching. The former struggle with object symmetries, while the latter fail in the absence of distinctive local features. To overcome these limitations, we propose a novel formulation of 6D pose estimation as a conditional flow matching problem in $\\mathbb{R}^3$. We introduce Flose, a generative method that infers object poses via a denoising process conditioned on local features. While prior approaches based on conditional flow matching perform denoising solely based on geometric guidance, Flose integrates appearance-based semantic features to mitigate ambiguities caused by object symmetries. We further incorporate RANSAC-based registration to handle outliers. We validate Flose on five datasets from the established BOP benchmark. Flose outperforms prior methods with an average improvement of +4.5 Average Recall. Project Website : https://tev-fbk.github.io/Flose/", "date": "2026-02-23", "categories": []}
{"id": "2602.19937", "title": "Learning Positive-Incentive Point Sampling in Neural Implicit Fields for Object Pose Estimation", "abstract": "Learning neural implicit fields of 3D shapes is a rapidly emerging field that enables shape representation at arbitrary resolutions. Due to the flexibility, neural implicit fields have succeeded in many research areas, including shape reconstruction, novel view image synthesis, and more recently, object pose estimation. Neural implicit fields enable learning dense correspondences between the camera space and the object's canonical space-including unobserved regions in camera space-significantly boosting object pose estimation performance in challenging scenarios like highly occluded objects and novel shapes. Despite progress, predicting canonical coordinates for unobserved camera-space regions remains challenging due to the lack of direct observational signals. This necessitates heavy reliance on the model's generalization ability, resulting in high uncertainty. Consequently, densely sampling points across the entire camera space may yield inaccurate estimations that hinder the learning process and compromise performance. To alleviate this problem, we propose a method combining an SO(3)-equivariant convolutional implicit network and a positive-incentive point sampling (PIPS) strategy. The SO(3)-equivariant convolutional implicit network estimates point-level attributes with SO(3)-equivariance at arbitrary query locations, demonstrating superior performance compared to most existing baselines. The PIPS strategy dynamically determines sampling locations based on the input, thereby boosting the network's accuracy and training efficiency. Our method outperforms the state-of-the-art on three pose estimation datasets. Notably, it demonstrates significant improvements in challenging scenarios, such as objects captured with unseen pose, high occlusion, novel geometry, and severe noise.", "date": "2026-02-23", "categories": []}
{"id": "2405.07801", "title": "Deep Learning-Based Object Pose Estimation: A Comprehensive Survey", "abstract": "Object pose estimation is a fundamental computer vision problem with broad applications in augmented reality and robotics. Over the past decade, deep learning models, due to their superior accuracy and robustness, have increasingly supplanted conventional algorithms reliant on engineered point pair features. Nevertheless, several challenges persist in contemporary methods, including their dependency on labeled training data, model compactness, robustness under challenging conditions, and their ability to generalize to novel unseen objects. A recent survey discussing the progress made on different aspects of this area, outstanding challenges, and promising future directions, is missing. To fill this gap, we discuss the recent advances in deep learning-based object pose estimation, covering all three formulations of the problem, \\emph{i.e.}, instance-level, category-level, and unseen object pose estimation. Our survey also covers multiple input data modalities, degrees-of-freedom of output poses, object properties, and downstream tasks, providing the readers with a holistic understanding of this field. Additionally, it discusses training paradigms of different domains, inference modes, application areas, evaluation metrics, and benchmark datasets, as well as reports the performance of current state-of-the-art methods on these benchmarks, thereby facilitating the readers in selecting the most suitable method for their application. Finally, the survey identifies key challenges, reviews the prevailing trends along with their pros and cons, and identifies promising directions for future research. We also keep tracing the latest works at https://github.com/CNJianLiu/Awesome-Object-Pose-Estimation.", "date": "2024-05-13", "categories": []}
{"id": "2503.11692", "title": "FloPE: Flower Pose Estimation for Precision Pollination", "abstract": "This study presents Flower Pose Estimation (FloPE), a real-time flower pose estimation framework for computationally constrained robotic pollination systems. Robotic pollination has been proposed to supplement natural pollination to ensure global food security due to the decreased population of natural pollinators. However, flower pose estimation for pollination is challenging due to natural variability, flower clusters, and high accuracy demands due to the flowers' fragility when pollinating. This method leverages 3D Gaussian Splatting to generate photorealistic synthetic datasets with precise pose annotations, enabling effective knowledge distillation from a high-capacity teacher model to a lightweight student model for efficient inference. The approach was evaluated on both single and multi-arm robotic platforms, achieving a mean pose estimation error of 0.6 cm and 19.14 degrees within a low computational cost. Our experiments validate the effectiveness of FloPE, achieving up to 78.75% pollination success rate and outperforming prior robotic pollination techniques.", "date": "2025-03-08", "categories": []}
{"id": "2602.06211", "title": "DroneKey++: A Size Prior-free Method and New Benchmark for Drone 3D Pose Estimation from Sequential Images", "abstract": "Accurate 3D pose estimation of drones is essential for security and surveillance systems. However, existing methods often rely on prior drone information such as physical sizes or 3D meshes. At the same time, current datasets are small-scale, limited to single models, and collected under constrained environments, which makes reliable validation of generalization difficult. We present DroneKey++, a prior-free framework that jointly performs keypoint detection, drone classification, and 3D pose estimation. The framework employs a keypoint encoder for simultaneous keypoint detection and classification, and a pose decoder that estimates 3D pose using ray-based geometric reasoning and class embeddings. To address dataset limitations, we construct 6DroneSyn, a large-scale synthetic benchmark with over 50K images covering 7 drone models and 88 outdoor backgrounds, generated using 360-degree panoramic synthesis. Experiments show that DroneKey++ achieves MAE 17.34 deg and MedAE 17.1 deg for rotation, MAE 0.135 m and MedAE 0.242 m for translation, with inference speeds of 19.25 FPS (CPU) and 414.07 FPS (GPU), demonstrating both strong generalization across drone models and suitability for real-time applications. The dataset is publicly available.", "date": "2026-02-05", "categories": []}
{"id": "2602.05257", "title": "RFM-Pose:Reinforcement-Guided Flow Matching for Fast Category-Level 6D Pose Estimation", "abstract": "Object pose estimation is a fundamental problem in computer vision and plays a critical role in virtual reality and embodied intelligence, where agents must understand and interact with objects in 3D space. Recently, score based generative models have to some extent solved the rotational symmetry ambiguity problem in category level pose estimation, but their efficiency remains limited by the high sampling cost of score-based diffusion. In this work, we propose a new framework, RFM-Pose, that accelerates category-level 6D object pose generation while actively evaluating sampled hypotheses. To improve sampling efficiency, we adopt a flow-matching generative model and generate pose candidates along an optimal transport path from a simple prior to the pose distribution. To further refine these candidates, we cast the flow-matching sampling process as a Markov decision process and apply proximal policy optimization to fine-tune the sampling policy. In particular, we interpret the flow field as a learnable policy and map an estimator to a value network, enabling joint optimization of pose generation and hypothesis scoring within a reinforcement learning framework. Experiments on the REAL275 benchmark demonstrate that RFM-Pose achieves favorable performance while significantly reducing computational cost. Moreover, similar to prior work, our approach can be readily adapted to object pose tracking and attains competitive results in this setting.", "date": "2026-02-05", "categories": []}
{"id": "2602.05555", "title": "IndustryShapes: An RGB-D Benchmark dataset for 6D object pose estimation of industrial assembly components and tools", "abstract": "We introduce IndustryShapes, a new RGB-D benchmark dataset of industrial tools and components, designed for both instance-level and novel object 6D pose estimation approaches. The dataset provides a realistic and application-relevant testbed for benchmarking these methods in the context of industrial robotics bridging the gap between lab-based research and deployment in real-world manufacturing scenarios. Unlike many previous datasets that focus on household or consumer products or use synthetic, clean tabletop datasets, or objects captured solely in controlled lab environments, IndustryShapes introduces five new object types with challenging properties, also captured in realistic industrial assembly settings. The dataset has diverse complexity, from simple to more challenging scenes, with single and multiple objects, including scenes with multiple instances of the same object and it is organized in two parts: the classic set and the extended set. The classic set includes a total of 4,6k images and 6k annotated poses. The extended set introduces additional data modalities to support the evaluation of model-free and sequence-based approaches. To the best of our knowledge, IndustryShapes is the first dataset to offer RGB-D static onboarding sequences. We further evaluate the dataset on a representative set of state-of-the art methods for instance-based and novel object 6D pose estimation, including also object detection, segmentation, showing that there is room for improvement in this domain. The dataset page can be found in https://pose-lab.github.io/IndustryShapes.", "date": "2026-02-05", "categories": []}
{"id": "2601.12090", "title": "Detecting 3D Line Segments for 6DoF Pose Estimation with Limited Data", "abstract": "The task of 6DoF object pose estimation is one of the fundamental problems of 3D vision with many practical applications such as industrial automation. Traditional deep learning approaches for this task often require extensive training data or CAD models, limiting their application in real-world industrial settings where data is scarce and object instances vary. We propose a novel method for 6DoF pose estimation focused specifically on bins used in industrial settings. We exploit the cuboid geometry of bins by first detecting intermediate 3D line segments corresponding to their top edges. Our approach extends the 2D line segment detection network LeTR to operate on structured point cloud data. The detected 3D line segments are then processed using a simple geometric procedure to robustly determine the bin's 6DoF pose. To evaluate our method, we extend an existing dataset with a newly collected and annotated dataset, which we make publicly available. We show that incorporating synthetic training data significantly improves pose estimation accuracy on real scans. Moreover, we show that our method significantly outperforms current state-of-the-art 6DoF pose estimation methods in terms of the pose accuracy (3 cm translation error, 8.2$^\\circ$ rotation error) while not requiring instance-specific CAD models during inference.", "date": "2026-01-17", "categories": []}
{"id": "2406.09260", "title": "Deep Transformer Network for Monocular Pose Estimation of Shipborne Unmanned Aerial Vehicle", "abstract": "This paper introduces a deep transformer network for estimating the relative 6D pose of a Unmanned Aerial Vehicle (UAV) with respect to a ship using monocular images. A synthetic dataset of ship images is created and annotated with 2D keypoints of multiple ship parts. A Transformer Neural Network model is trained to detect these keypoints and estimate the 6D pose of each part. The estimates are integrated using Bayesian fusion. The model is tested on synthetic data and in-situ flight experiments, demonstrating robustness and accuracy in various lighting conditions. The position estimation error is approximately 0.8\\% and 1.0\\% of the distance to the ship for the synthetic data and the flight experiments, respectively. The method has potential applications for ship-based autonomous UAV landing and navigation.", "date": "2024-06-13", "categories": []}
{"id": "2407.12138", "title": "Monocular pose estimation of articulated open surgery tools -- in the wild", "abstract": "This work presents a framework for monocular 6D pose estimation of surgical instruments in open surgery, addressing challenges such as object articulations, specularity, occlusions, and synthetic-to-real domain adaptation. The proposed approach consists of three main components: $(1)$ synthetic data generation pipeline that incorporates 3D scanning of surgical tools with articulation rigging and physically-based rendering; $(2)$ a tailored pose estimation framework combining tool detection with pose and articulation estimation; and $(3)$ a training strategy on synthetic and real unannotated video data, employing domain adaptation with automatically generated pseudo-labels. Evaluations conducted on real data of open surgery demonstrate the good performance and real-world applicability of the proposed framework, highlighting its potential for integration into medical augmented reality and robotic systems. The approach eliminates the need for extensive manual annotation of real surgical data.", "date": "2024-07-16", "categories": []}
{"id": "2503.00051", "title": "Correspondence-Free Pose Estimation with Patterns: A Unified Approach for Multi-Dimensional Vision", "abstract": "6D pose estimation is a central problem in robot vision. Compared with pose estimation based on point correspondences or its robust versions, correspondence-free methods are often more flexible. However, existing correspondence-free methods often rely on feature representation alignment or end-to-end regression. For such a purpose, a new correspondence-free pose estimation method and its practical algorithms are proposed, whose key idea is the elimination of unknowns by process of addition to separate the pose estimation from correspondence. By taking the considered point sets as patterns, feature functions used to describe these patterns are introduced to establish a sufficient number of equations for optimization. The proposed method is applicable to nonlinear transformations such as perspective projection and can cover various pose estimations from 3D-to-3D points, 3D-to-2D points, and 2D-to-2D points. Experimental results on both simulation and actual data are presented to demonstrate the effectiveness of the proposed method.", "date": "2025-02-26", "categories": []}
{"id": "2508.03618", "title": "FPG-NAS: FLOPs-Aware Gated Differentiable Neural Architecture Search for Efficient 6DoF Pose Estimation", "abstract": "We introduce FPG-NAS, a FLOPs-aware Gated Differentiable Neural Architecture Search framework for efficient 6DoF object pose estimation. Estimating 3D rotation and translation from a single image has been widely investigated yet remains computationally demanding, limiting applicability in resource-constrained scenarios. FPG-NAS addresses this by proposing a specialized differentiable NAS approach for 6DoF pose estimation, featuring a task-specific search space and a differentiable gating mechanism that enables discrete multi-candidate operator selection, thus improving architectural diversity. Additionally, a FLOPs regularization term ensures a balanced trade-off between accuracy and efficiency. The framework explores a vast search space of approximately 10\\textsuperscript{92} possible architectures. Experiments on the LINEMOD and SPEED+ datasets demonstrate that FPG-NAS-derived models outperform previous methods under strict FLOPs constraints. To the best of our knowledge, FPG-NAS is the first differentiable NAS framework specifically designed for 6DoF object pose estimation.", "date": "2025-08-05", "categories": []}
{"id": "2601.06883", "title": "MixRI: Mixing Features of Reference Images for Novel Object Pose Estimation", "abstract": "We present MixRI, a lightweight network that solves the CAD-based novel object pose estimation problem in RGB images. It can be instantly applied to a novel object at test time without finetuning. We design our network to meet the demands of real-world applications, emphasizing reduced memory requirements and fast inference time. Unlike existing works that utilize many reference images and have large network parameters, we directly match points based on the multi-view information between the query and reference images with a lightweight network. Thanks to our reference image fusion strategy, we significantly decrease the number of reference images, thus decreasing the time needed to process these images and the memory required to store them. Furthermore, with our lightweight network, our method requires less inference time. Though with fewer reference images, experiments on seven core datasets in the BOP challenge show that our method achieves comparable results with other methods that require more reference images and larger network parameters.", "date": "2026-01-11", "categories": []}
{"id": "2601.05573", "title": "Orient Anything V2: Unifying Orientation and Rotation Understanding", "abstract": "This work presents Orient Anything V2, an enhanced foundation model for unified understanding of object 3D orientation and rotation from single or paired images. Building upon Orient Anything V1, which defines orientation via a single unique front face, V2 extends this capability to handle objects with diverse rotational symmetries and directly estimate relative rotations. These improvements are enabled by four key innovations: 1) Scalable 3D assets synthesized by generative models, ensuring broad category coverage and balanced data distribution; 2) An efficient, model-in-the-loop annotation system that robustly identifies 0 to N valid front faces for each object; 3) A symmetry-aware, periodic distribution fitting objective that captures all plausible front-facing orientations, effectively modeling object rotational symmetry; 4) A multi-frame architecture that directly predicts relative object rotations. Extensive experiments show that Orient Anything V2 achieves state-of-the-art zero-shot performance on orientation estimation, 6DoF pose estimation, and object symmetry recognition across 11 widely used benchmarks. The model demonstrates strong generalization, significantly broadening the applicability of orientation estimation in diverse downstream tasks.", "date": "2026-01-09", "categories": []}
{"id": "2408.17297", "title": "BOP-Distrib: Revisiting 6D Pose Estimation Benchmarks for Better Evaluation under Visual Ambiguities", "abstract": "6D pose estimation aims at determining the object pose that best explains the camera observation. The unique solution for non-ambiguous objects can turn into a multi-modal pose distribution for symmetrical objects or when occlusions of symmetry-breaking elements happen, depending on the viewpoint. Currently, 6D pose estimation methods are benchmarked on datasets that consider, for their ground truth annotations, visual ambiguities as only related to global object symmetries, whereas they should be defined per-image to account for the camera viewpoint. We thus first propose an automatic method to re-annotate those datasets with a 6D pose distribution specific to each image, taking into account the object surface visibility in the image to correctly determine the visual ambiguities. Second, given this improved ground truth, we re-evaluate the state-of-the-art single pose methods and show that this greatly modifies the ranking of these methods. Third, as some recent works focus on estimating the complete set of solutions, we derive a precision/recall formulation to evaluate them against our image-wise distribution ground truth, making it the first benchmark for pose distribution methods on real images.", "date": "2024-08-30", "categories": []}
{"id": "2501.01993", "title": "A Novel Convolution and Attention Mechanism-based Model for 6D Object Pose Estimation", "abstract": "This paper proposes PoseLecTr, a graph-based encoder-decoder framework that integrates a novel Legendre convolution with attention mechanisms for six-degree-of-freedom (6-DOF) object pose estimation from monocular RGB images. Conventional learning-based approaches predominantly rely on grid-structured convolutions, which can limit their ability to model higher-order and long-range dependencies among image features, especially in cluttered or occluded scenes. PoseLecTr addresses this limitation by constructing a graph representation from image features, where spatial relationships are explicitly modeled through graph connectivity. The proposed framework incorporates a Legendre convolution layer to improve numerical stability in graph convolution, together with spatial-attention and self-attention distillation to enhance feature selection. Experiments conducted on the LINEMOD, Occluded LINEMOD, and YCB-VIDEO datasets demonstrate that our method achieves competitive performance and shows consistent improvements across a wide range of objects and scene complexities.", "date": "2024-12-31", "categories": []}
{"id": "2512.22979", "title": "PoseStreamer: A Multi-modal Framework for 3D Tracking of Unseen Moving Objects", "abstract": "Six degree of freedom (6DoF) pose estimation for novel objects is a critical task in computer vision, yet it faces significant challenges in high-speed and low-light scenarios where standard RGB cameras suffer from motion blur. While event cameras offer a promising solution due to their high temporal resolution, current 6DoF pose estimation methods typically yield suboptimal performance in high-speed object moving scenarios. To address this gap, we propose PoseStreamer, a robust multi-modal 6DoF pose estimation framework designed specifically on high-speed moving scenarios. Our approach integrates three core components: an Adaptive Pose Memory Queue that utilizes historical orientation cues for temporal consistency, an Object-centric 2D Tracker that provides strong 2D priors to boost 3D center recall, and a Ray Pose Filter for geometric refinement along camera rays. Furthermore, we introduce MoCapCube6D, a novel multi-modal dataset constructed to benchmark performance under rapid motion. Extensive experiments demonstrate that PoseStreamer not only achieves superior accuracy in high-speed moving scenarios, but also exhibits strong generalizability as a template-free framework for unseen moving objects.", "date": "2025-12-28", "categories": []}
{"id": "2512.22581", "title": "KV-Tracker: Real-Time Pose Tracking with Transformers", "abstract": "Multi-view 3D geometry networks offer a powerful prior but are prohibitively slow for real-time applications. We propose a novel way to adapt them for online use, enabling real-time 6-DoF pose tracking and online reconstruction of objects and scenes from monocular RGB videos. Our method rapidly selects and manages a set of images as keyframes to map a scene or object via $\\pi^3$ with full bidirectional attention. We then cache the global self-attention block's key-value (KV) pairs and use them as the sole scene representation for online tracking. This allows for up to $15\\times$ speedup during inference without the fear of drift or catastrophic forgetting. Our caching strategy is model-agnostic and can be applied to other off-the-shelf multi-view networks without retraining. We demonstrate KV-Tracker on both scene-level tracking and the more challenging task of on-the-fly object tracking and reconstruction without depth measurements or object priors. Experiments on the TUM RGB-D, 7-Scenes, Arctic and OnePose datasets show the strong performance of our system while maintaining high frame-rates up to ${\\sim}27$ FPS.", "date": "2025-12-27", "categories": []}
{"id": "2512.21053", "title": "Optical Flow-Guided 6DoF Object Pose Tracking with an Event Camera", "abstract": "Object pose tracking is one of the pivotal technologies in multimedia, attracting ever-growing attention in recent years. Existing methods employing traditional cameras encounter numerous challenges such as motion blur, sensor noise, partial occlusion, and changing lighting conditions. The emerging bio-inspired sensors, particularly event cameras, possess advantages such as high dynamic range and low latency, which hold the potential to address the aforementioned challenges. In this work, we present an optical flow-guided 6DoF object pose tracking method with an event camera. A 2D-3D hybrid feature extraction strategy is firstly utilized to detect corners and edges from events and object models, which characterizes object motion precisely. Then, we search for the optical flow of corners by maximizing the event-associated probability within a spatio-temporal window, and establish the correlation between corners and edges guided by optical flow. Furthermore, by minimizing the distances between corners and edges, the 6DoF object pose is iteratively optimized to achieve continuous pose tracking. Experimental results of both simulated and real events demonstrate that our methods outperform event-based state-of-the-art methods in terms of both accuracy and robustness.", "date": "2025-12-24", "categories": []}
{"id": "2512.18784", "title": "Eff-GRot: Efficient and Generalizable Rotation Estimation with Transformers", "abstract": "We introduce Eff-GRot, an approach for efficient and generalizable rotation estimation from RGB images. Given a query image and a set of reference images with known orientations, our method directly predicts the object's rotation in a single forward pass, without requiring object- or category-specific training. At the core of our framework is a transformer that performs a comparison in the latent space, jointly processing rotation-aware representations from multiple references alongside a query. This design enables a favorable balance between accuracy and computational efficiency while remaining simple, scalable, and fully end-to-end. Experimental results show that Eff-GRot offers a promising direction toward more efficient rotation estimation, particularly in latency-sensitive applications.", "date": "2025-12-21", "categories": []}
{"id": "2512.19058", "title": "6DAttack: Backdoor Attacks in the 6DoF Pose Estimation", "abstract": "Deep learning advances have enabled accurate six-degree-of-freedom (6DoF) object pose estimation, widely used in robotics, AR/VR, and autonomous systems. However, backdoor attacks pose significant security risks. While most research focuses on 2D vision, 6DoF pose estimation remains largely unexplored. Unlike traditional backdoors that only change classes, 6DoF attacks must control continuous parameters like translation and rotation, rendering 2D methods inapplicable. We propose 6DAttack, a framework using 3D object triggers to induce controlled erroneous poses while maintaining normal behavior. Evaluations on PVNet, DenseFusion, and PoseDiffusion across LINEMOD, YCB-Video, and CO3D show high attack success rates (ASRs) without compromising clean performance. Backdoored models achieve up to 100% clean ADD accuracy and 100% ASR, with triggered samples reaching 97.70% ADD-P. Furthermore, a representative defense remains ineffective. Our findings reveal a serious, underexplored threat to 6DoF pose estimation.", "date": "2025-12-22", "categories": []}
{"id": "2503.13053", "title": "Uncertainty-Aware Knowledge Distillation for Compact and Efficient 6DoF Pose Estimation", "abstract": "Compact and efficient 6DoF object pose estimation is crucial in applications such as robotics, augmented reality, and space autonomous navigation systems, where lightweight models are critical for real-time accurate performance. This paper introduces a novel uncertainty-aware end-to-end Knowledge Distillation (KD) framework focused on keypoint-based 6DoF pose estimation. Keypoints predicted by a large teacher model exhibit varying levels of uncertainty that can be exploited within the distillation process to enhance the accuracy of the student model while ensuring its compactness. To this end, we propose a distillation strategy that aligns the student and teacher predictions by adjusting the knowledge transfer based on the uncertainty associated with each teacher keypoint prediction. Additionally, the proposed KD leverages this uncertainty-aware alignment of keypoints to transfer the knowledge at key locations of their respective feature maps. Experiments on the widely-used LINEMOD benchmark demonstrate the effectiveness of our method, achieving superior 6DoF object pose estimation with lightweight models compared to state-of-the-art approaches. Further validation on the SPEED+ dataset for spacecraft pose estimation highlights the robustness of our approach under diverse 6DoF pose estimation scenarios.", "date": "2025-03-17", "categories": []}
{"id": "2503.03726", "title": "Active 6D Pose Estimation for Textureless Objects using Multi-View RGB Frames", "abstract": "Estimating the 6D pose of textureless objects from RGB images is an important problem in robotics. Due to appearance ambiguities, rotational symmetries, and severe occlusions, single-view based 6D pose estimators are still unable to handle a wide range of objects, motivating research towards multi-view pose estimation and next-best-view prediction that addresses these limitations. In this work, we propose a comprehensive active perception framework for estimating the 6D poses of textureless objects using only RGB images. Our approach is built upon a key idea: decoupling the 6D pose estimation into a two-step sequential process can greatly improve both accuracy and efficiency. First, we estimate the 3D translation of each object, resolving scale and depth ambiguities inherent to RGB images. These estimates are then used to simplify the subsequent task of determining the 3D orientation, which we achieve through canonical scale template matching. Building on this formulation, we then introduce an active perception strategy that predicts the next best camera viewpoint to capture an RGB image, effectively reducing object pose uncertainty and enhancing pose accuracy. We evaluate our method on the public ROBI and TOD datasets, as well as on our reconstructed transparent object dataset, T-ROBI. Under the same camera viewpoints, our multi-view pose estimation significantly outperforms state-of-the-art approaches. Furthermore, by leveraging our next-best-view strategy, our approach achieves high pose accuracy with fewer viewpoints than heuristic-based policies across all evaluated datasets. The accompanying video and T-ROBI dataset will be released on our project page: https://trailab.github.io/ActiveODPE.", "date": "2025-03-05", "categories": []}
{"id": "2504.15134", "title": "Instance-Adaptive Keypoint Learning with Local-to-Global Geometric Aggregation for Category-Level Object Pose Estimation", "abstract": "Category-level object pose estimation aims to predict the 6D pose and size of previously unseen instances from predefined categories, requiring strong generalization across diverse object instances. Although many previous methods attempt to mitigate intra-class variations, they often struggle with instances exhibiting complex geometries or significant deviations from canonical shapes. To address this issue, we propose INKL-Pose, a novel category-level object pose estimation framework that enables INstance-adaptive Keypoint Learning with local-to-global geometric aggregation. Specifically, our method first predicts semantically consistent and geometrically informative keypoints using an Instance-Adaptive Keypoint Detector, then refines them: (1) a Local Keypoint Feature Aggregator capturing fine-grained geometries, and (2) a Global Keypoint Feature Aggregator using bidirectional Mamba for structural consistency. To enable bidirectional modeling in Mamba, we introduce a simple yet effective Feature Sequence Flipping strategy that preserves spatial coherence while constructing backward feature sequence. Additionally, we design a surface loss and a separation loss to encourage uniform coverage and spatial diversity in keypoint distribution. The resulting keypoints are mapped to a canonical space for 6D pose and size regression. Extensive experiments on CAMERA25, REAL275, and HouseCat6D show that INKL-Pose achieves state-of-the-art performance with 16.7M parameters and runs at 36 FPS on an NVIDIA RTX 4090D GPU.", "date": "2025-04-21", "categories": []}
{"id": "2511.12919", "title": "CoordAR: One-Reference 6D Pose Estimation of Novel Objects via Autoregressive Coordinate Map Generation", "abstract": "Object 6D pose estimation, a crucial task for robotics and augmented reality applications, becomes particularly challenging when dealing with novel objects whose 3D models are not readily available. To reduce dependency on 3D models, recent studies have explored one-reference-based pose estimation, which requires only a single reference view instead of a complete 3D model. However, existing methods that rely on real-valued coordinate regression suffer from limited global consistency due to the local nature of convolutional architectures and face challenges in symmetric or occluded scenarios owing to a lack of uncertainty modeling. We present CoordAR, a novel autoregressive framework for one-reference 6D pose estimation of unseen objects. CoordAR formulates 3D-3D correspondences between the reference and query views as a map of discrete tokens, which is obtained in an autoregressive and probabilistic manner. To enable accurate correspondence regression, CoordAR introduces 1) a novel coordinate map tokenization that enables probabilistic prediction over discretized 3D space; 2) a modality-decoupled encoding strategy that separately encodes RGB appearance and coordinate cues; and 3) an autoregressive transformer decoder conditioned on both position-aligned query features and the partially generated token sequence. With these novel mechanisms, CoordAR significantly outperforms existing methods on multiple benchmarks and demonstrates strong robustness to symmetry, occlusion, and other challenges in real-world tests.", "date": "2025-11-17", "categories": []}
{"id": "2511.13102", "title": "CapeNext: Rethinking and Refining Dynamic Support Information for Category-Agnostic Pose Estimation", "abstract": "Recent research in Category-Agnostic Pose Estimation (CAPE) has adopted fixed textual keypoint description as semantic prior for two-stage pose matching frameworks. While this paradigm enhances robustness and flexibility by disentangling the dependency of support images, our critical analysis reveals two inherent limitations of static joint embedding: (1) polysemy-induced cross-category ambiguity during the matching process(e.g., the concept \"leg\" exhibiting divergent visual manifestations across humans and furniture), and (2) insufficient discriminability for fine-grained intra-category variations (e.g., posture and fur discrepancies between a sleeping white cat and a standing black cat). To overcome these challenges, we propose a new framework that innovatively integrates hierarchical cross-modal interaction with dual-stream feature refinement, enhancing the joint embedding with both class-level and instance-specific cues from textual description and specific images. Experiments on the MP-100 dataset demonstrate that, regardless of the network backbone, CapeNext consistently outperforms state-of-the-art CAPE methods by a large margin.", "date": "2025-11-17", "categories": []}
{"id": "2512.13007", "title": "Light Field Based 6DoF Tracking of Previously Unobserved Objects", "abstract": "Object tracking is an important step in robotics and reautonomous driving pipelines, which has to generalize to previously unseen and complex objects. Existing high-performing methods often rely on pre-captured object views to build explicit reference models, which restricts them to a fixed set of known objects. However, such reference models can struggle with visually complex appearance, reducing the quality of tracking. In this work, we introduce an object tracking method based on light field images that does not depend on a pre-trained model, while being robust to complex visual behavior, such as reflections. We extract semantic and geometric features from light field inputs using vision foundation models and convert them into view-dependent Gaussian splats. These splats serve as a unified object representation, supporting differentiable rendering and pose optimization. We further introduce a light field object tracking dataset containing challenging reflective objects with precise ground truth poses. Experiments demonstrate that our method is competitive with state-of-the-art model-based trackers in these difficult cases, paving the way toward universal object tracking in robotic systems. Code/data available at https://github.com/nagonch/LiFT-6DoF.", "date": "2025-12-15", "categories": []}
{"id": "2512.10251", "title": "THE-Pose: Topological Prior with Hybrid Graph Fusion for Estimating Category-Level 6D Object Pose", "abstract": "Category-level object pose estimation requires both global context and local structure to ensure robustness against intra-class variations. However, 3D graph convolution (3D-GC) methods only focus on local geometry and depth information, making them vulnerable to complex objects and visual ambiguities. To address this, we present THE-Pose, a novel category-level 6D pose estimation framework that leverages a topological prior via surface embedding and hybrid graph fusion. Specifically, we extract consistent and invariant topological features from the image domain, effectively overcoming the limitations inherent in existing 3D-GC based methods. Our Hybrid Graph Fusion (HGF) module adaptively integrates the topological features with point-cloud features, seamlessly bridging 2D image context and 3D geometric structure. These fused features ensure stability for unseen or complicated objects, even under significant occlusions. Extensive experiments on the REAL275 dataset show that THE-Pose achieves a 35.8% improvement over the 3D-GC baseline (HS-Pose) and surpasses the previous state-of-the-art by 7.2% across all key metrics. The code is avaialbe on https://github.com/EHxxx/THE-Pose", "date": "2025-12-11", "categories": []}
{"id": "2512.10674", "title": "Geo6DPose: Fast Zero-Shot 6D Object Pose Estimation via Geometry-Filtered Feature Matching", "abstract": "Recent progress in zero-shot 6D object pose estimation has been driven largely by large-scale models and cloud-based inference. However, these approaches often introduce high latency, elevated energy consumption, and deployment risks related to connectivity, cost, and data governance; factors that conflict with the practical constraints of real-world robotics, where compute is limited and on-device inference is frequently required. We introduce Geo6DPose, a lightweight, fully local, and training-free pipeline for zero-shot 6D pose estimation that trades model scale for geometric reliability. Our method combines foundation model visual features with a geometric filtering strategy: Similarity maps are computed between onboarded template DINO descriptors and scene patches, and mutual correspondences are established by projecting scene patch centers to 3D and template descriptors to the object model coordinate system. Final poses are recovered via correspondence-driven RANSAC and ranked using a weighted geometric alignment metric that jointly accounts for reprojection consistency and spatial support, improving robustness to noise, clutter, and partial visibility. Geo6DPose achieves sub-second inference on a single commodity GPU while matching the average recall of significantly larger zero-shot baselines (53.7 AR, 1.08 FPS). It requires no training, fine-tuning, or network access, and remains compatible with evolving foundation backbones, advancing practical, fully local 6D perception for robotic deployment.", "date": "2025-12-11", "categories": []}
{"id": "2512.09792", "title": "FastPose-ViT: A Vision Transformer for Real-Time Spacecraft Pose Estimation", "abstract": "Estimating the 6-degrees-of-freedom (6DoF) pose of a spacecraft from a single image is critical for autonomous operations like in-orbit servicing and space debris removal. Existing state-of-the-art methods often rely on iterative Perspective-n-Point (PnP)-based algorithms, which are computationally intensive and ill-suited for real-time deployment on resource-constrained edge devices. To overcome these limitations, we propose FastPose-ViT, a Vision Transformer (ViT)-based architecture that directly regresses the 6DoF pose. Our approach processes cropped images from object bounding boxes and introduces a novel mathematical formalism to map these localized predictions back to the full-image scale. This formalism is derived from the principles of projective geometry and the concept of \"apparent rotation\", where the model predicts an apparent rotation matrix that is then corrected to find the true orientation. We demonstrate that our method outperforms other non-PnP strategies and achieves performance competitive with state-of-the-art PnP-based techniques on the SPEED dataset. Furthermore, we validate our model's suitability for real-world space missions by quantizing it and deploying it on power-constrained edge hardware. On the NVIDIA Jetson Orin Nano, our end-to-end pipeline achieves a latency of ~75 ms per frame under sequential execution, and a non-blocking throughput of up to 33 FPS when stages are scheduled concurrently.", "date": "2025-12-10", "categories": []}
{"id": "2512.07215", "title": "VFM-VLM: Vision Foundation Model and Vision Language Model based Visual Comparison for 3D Pose Estimation", "abstract": "Vision Foundation Models (VFMs) and Vision Language Models (VLMs) have revolutionized computer vision by providing rich semantic and geometric representations. This paper presents a comprehensive visual comparison between CLIP based and DINOv2 based approaches for 3D pose estimation in hand object grasping scenarios. We evaluate both models on the task of 6D object pose estimation and demonstrate their complementary strengths: CLIP excels in semantic understanding through language grounding, while DINOv2 provides superior dense geometric features. Through extensive experiments on benchmark datasets, we show that CLIP based methods achieve better semantic consistency, while DINOv2 based approaches demonstrate competitive performance with enhanced geometric precision. Our analysis provides insights for selecting appropriate vision models for robotic manipulation and grasping, picking applications.", "date": "2025-12-08", "categories": []}
{"id": "2512.08430", "title": "SDT-6D: Fully Sparse Depth-Transformer for Staged End-to-End 6D Pose Estimation in Industrial Multi-View Bin Picking", "abstract": "Accurately recovering 6D poses in densely packed industrial bin-picking environments remain a serious challenge, owing to occlusions, reflections, and textureless parts. We introduce a holistic depth-only 6D pose estimation approach that fuses multi-view depth maps into either a fine-grained 3D point cloud in its vanilla version, or a sparse Truncated Signed Distance Field (TSDF). At the core of our framework lies a staged heatmap mechanism that yields scene-adaptive attention priors across different resolutions, steering computation toward foreground regions, thus keeping memory requirements at high resolutions feasible. Along, we propose a density-aware sparse transformer block that dynamically attends to (self-) occlusions and the non-uniform distribution of 3D data. While sparse 3D approaches has proven effective for long-range perception, its potential in close-range robotic applications remains underexplored. Our framework operates fully sparse, enabling high-resolution volumetric representations to capture fine geometric details crucial for accurate pose estimation in clutter. Our method processes the entire scene integrally, predicting the 6D pose via a novel per-voxel voting strategy, allowing simultaneous pose predictions for an arbitrary number of target objects. We validate our method on the recently published IPD and MV-YCB multi-view datasets, demonstrating competitive performance in heavily cluttered industrial and household bin picking scenarios.", "date": "2025-12-09", "categories": []}
{"id": "2409.11661", "title": "Bridging the Domain Gap for Flight-Ready Spaceborne Vision", "abstract": "This work presents Spacecraft Pose Network v3 (SPNv3), a Neural Network (NN) for monocular pose estimation of a known, non-cooperative target spacecraft. SPNv3 is designed and trained to be computationally efficient while providing robustness to spaceborne images that have not been observed during offline training and validation on the ground. These characteristics are essential to deploying NNs on space-grade edge devices. They are achieved through careful NN design choices, and an extensive trade-off analysis reveals features such as data augmentation, transfer learning and vision transformer architecture as a few of those that contribute to simultaneously maximizing robustness and minimizing computational overhead. Experiments demonstrate that the final SPNv3 can achieve state-of-the-art pose accuracy on hardware-in-the-loop images from a robotic testbed while having trained exclusively on computer-generated synthetic images, effectively bridging the domain gap between synthetic and real imagery. At the same time, SPNv3 runs well above the update frequency of modern satellite navigation filters when tested on a representative graphical processing unit system with flight heritage. Overall, SPNv3 is an efficient, flight-ready NN model readily applicable to close-range rendezvous and proximity operations with target resident space objects.", "date": "2024-09-18", "categories": []}
{"id": "2512.07211", "title": "Object Pose Distribution Estimation for Determining Revolution and Reflection Uncertainty in Point Clouds", "abstract": "Object pose estimation is crucial to robotic perception and typically provides a single-pose estimate. However, a single estimate cannot capture pose uncertainty deriving from visual ambiguity, which can lead to unreliable behavior. Existing pose distribution methods rely heavily on color information, often unavailable in industrial settings.\n  We propose a novel neural network-based method for estimating object pose uncertainty using only 3D colorless data. To the best of our knowledge, this is the first approach that leverages deep learning for pose distribution estimation without relying on RGB input. We validate our method in a real-world bin picking scenario with objects of varying geometric ambiguity. Our current implementation focuses on symmetries in reflection and revolution, but the framework is extendable to full SE(3) pose distribution estimation. Source code available at opde3d.github.io", "date": "2025-12-08", "categories": []}
{"id": "2408.16547", "title": "OP-Align: Object-level and Part-level Alignment for Self-supervised Category-level Articulated Object Pose Estimation", "abstract": "Category-level articulated object pose estimation focuses on the pose estimation of unknown articulated objects within known categories. Despite its significance, this task remains challenging due to the varying shapes and poses of objects, expensive dataset annotation costs, and complex real-world environments. In this paper, we propose a novel self-supervised approach that leverages a single-frame point cloud to solve this task. Our model consistently generates reconstruction with a canonical pose and joint state for the entire input object, and it estimates object-level poses that reduce overall pose variance and part-level poses that align each part of the input with its corresponding part of the reconstruction. Experimental results demonstrate that our approach significantly outperforms previous self-supervised methods and is comparable to the state-of-the-art supervised methods. To assess the performance of our model in real-world scenarios, we also introduce a new real-world articulated object benchmark dataset.", "date": "2024-08-29", "categories": []}
{"id": "2511.21666", "title": "Uncertainty Quantification for Visual Object Pose Estimation", "abstract": "Quantifying the uncertainty of an object's pose estimate is essential for robust control and planning. Although pose estimation is a well-studied robotics problem, attaching statistically rigorous uncertainty is not well understood without strict distributional assumptions. We develop distribution-free pose uncertainty bounds about a given pose estimate in the monocular setting. Our pose uncertainty only requires high probability noise bounds on pixel detections of 2D semantic keypoints on a known object. This noise model induces an implicit, non-convex set of pose uncertainty constraints. Our key contribution is SLUE (S-Lemma Uncertainty Estimation), a convex program to reduce this set to a single ellipsoidal uncertainty bound that is guaranteed to contain the true object pose with high probability. SLUE solves a relaxation of the minimum volume bounding ellipsoid problem inspired by the celebrated S-lemma. It requires no initial guess of the bound's shape or size and is guaranteed to contain the true object pose with high probability. For tighter uncertainty bounds at the same confidence, we extend SLUE to a sum-of-squares relaxation hierarchy which is guaranteed to converge to the minimum volume ellipsoidal uncertainty bound for a given set of keypoint constraints. We show this pose uncertainty bound can easily be projected to independent translation and axis-angle orientation bounds. We evaluate SLUE on two pose estimation datasets and a real-world drone tracking scenario. Compared to prior work, SLUE generates substantially smaller translation bounds and competitive orientation bounds. We release code at https://github.com/MIT-SPARK/PoseUncertaintySets.", "date": "2025-11-26", "categories": []}
{"id": "2511.19511", "title": "The Determinant Ratio Matrix Approach to Solving 3D Matching and 2D Orthographic Projection Alignment Tasks", "abstract": "Pose estimation is a general problem in computer vision with wide applications. The relative orientation of a 3D reference object can be determined from a 3D rotated version of that object, or from a projection of the rotated object to a 2D planar image. This projection can be a perspective projection (the PnP problem) or an orthographic projection (the OnP problem). We restrict our attention here to the OnP problem and the full 3D pose estimation task (the EnP problem). Here we solve the least squares systems for both the error-free EnP and OnP problems in terms of the determinant ratio matrix (DRaM) approach. The noisy-data case can be addressed with a straightforward rotation correction scheme. While the SVD and optimal quaternion eigensystem methods solve the noisy EnP 3D-3D alignment exactly, the noisy 3D-2D orthographic (OnP) task has no known comparable closed form, and can be solved by DRaM-class methods. We note that while previous similar work has been presented in the literature exploiting both the QR decomposition and the Moore-Penrose pseudoinverse transformations, here we place these methods in a larger context that has not previously been fully recognized in the absence of the corresponding DRaM solution. We term this class of solutions as the DRaM family, and conduct comparisons of the behavior of the families of solutions for the EnP and OnP rotation estimation problems. Overall, this work presents both a new solution to the 3D and 2D orthographic pose estimation problems and provides valuable insight into these classes of problems. With hindsight, we are able to show that our DRaM solutions to the exact EnP and OnP problems possess derivations that could have been discovered in the time of Gauss, and in fact generalize to all analogous N-dimensional Euclidean pose estimation problems.", "date": "2025-11-24", "categories": []}
{"id": "2503.18671", "title": "Structure-Aware Correspondence Learning for Relative Pose Estimation", "abstract": "Relative pose estimation provides a promising way for achieving object-agnostic pose estimation. Despite the success of existing 3D correspondence-based methods, the reliance on explicit feature matching suffers from small overlaps in visible regions and unreliable feature estimation for invisible regions. Inspired by humans' ability to assemble two object parts that have small or no overlapping regions by considering object structure, we propose a novel Structure-Aware Correspondence Learning method for Relative Pose Estimation, which consists of two key modules. First, a structure-aware keypoint extraction module is designed to locate a set of kepoints that can represent the structure of objects with different shapes and appearance, under the guidance of a keypoint based image reconstruction loss. Second, a structure-aware correspondence estimation module is designed to model the intra-image and inter-image relationships between keypoints to extract structure-aware features for correspondence estimation. By jointly leveraging these two modules, the proposed method can naturally estimate 3D-3D correspondences for unseen objects without explicit feature matching for precise relative pose estimation. Experimental results on the CO3D, Objaverse and LineMOD datasets demonstrate that the proposed method significantly outperforms prior methods, i.e., with 5.7{\\deg}reduction in mean angular error on the CO3D dataset.", "date": "2025-03-24", "categories": []}
{"id": "2511.15874", "title": "WALDO: Where Unseen Model-based 6D Pose Estimation Meets Occlusion", "abstract": "Accurate 6D object pose estimation is vital for robotics, augmented reality, and scene understanding. For seen objects, high accuracy is often attainable via per-object fine-tuning but generalizing to unseen objects remains a challenge. To address this problem, past arts assume access to CAD models at test time and typically follow a multi-stage pipeline to estimate poses: detect and segment the object, propose an initial pose, and then refine it. Under occlusion, however, the early-stage of such pipelines are prone to errors, which can propagate through the sequential processing, and consequently degrade the performance. To remedy this shortcoming, we propose four novel extensions to model-based 6D pose estimation methods: (i) a dynamic non-uniform dense sampling strategy that focuses computation on visible regions, reducing occlusion-induced errors; (ii) a multi-hypothesis inference mechanism that retains several confidence-ranked pose candidates, mitigating brittle single-path failures; (iii) iterative refinement to progressively improve pose accuracy; and (iv) series of occlusion-focused training augmentations that strengthen robustness and generalization. Furthermore, we propose a new weighted by visibility metric for evaluation under occlusion to minimize the bias in the existing protocols. Via extensive empirical evaluations, we show that our proposed approach achieves more than 5% improvement in accuracy on ICBIN and more than 2% on BOP dataset benchmarks, while achieving approximately 3 times faster inference.", "date": "2025-11-19", "categories": []}
{"id": "2511.15884", "title": "Box6D : Zero-shot Category-level 6D Pose Estimation of Warehouse Boxes", "abstract": "Accurate and efficient 6D pose estimation of novel objects under clutter and occlusion is critical for robotic manipulation across warehouse automation, bin picking, logistics, and e-commerce fulfillment. There are three main approaches in this domain; Model-based methods assume an exact CAD model at inference but require high-resolution meshes and transfer poorly to new environments; Model-free methods that rely on a few reference images or videos are more flexible, however often fail under challenging conditions; Category-level approaches aim to balance flexibility and accuracy but many are overly general and ignore environment and object priors, limiting their practicality in industrial settings.\n  To this end, we propose Box6d, a category-level 6D pose estimation method tailored for storage boxes in the warehouse context. From a single RGB-D observation, Box6D infers the dimensions of the boxes via a fast binary search and estimates poses using a category CAD template rather than instance-specific models. Suing a depth-based plausibility filter and early-stopping strategy, Box6D then rejects implausible hypotheses, lowering computational cost. We conduct evaluations on real-world storage scenarios and public benchmarks, and show that our approach delivers competitive or superior 6D pose precision while reducing inference time by approximately 76%.", "date": "2025-11-19", "categories": []}
{"id": "2511.12614", "title": "OPFormer: Object Pose Estimation leveraging foundation model with geometric encoding", "abstract": "We introduce a unified, end-to-end framework that seamlessly integrates object detection and pose estimation with a versatile onboarding process. Our pipeline begins with an onboarding stage that generates object representations from either traditional 3D CAD models or, in their absence, by rapidly reconstructing a high-fidelity neural representation (NeRF) from multi-view images. Given a test image, our system first employs the CNOS detector to localize target objects. For each detection, our novel pose estimation module, OPFormer, infers the precise 6D pose. The core of OPFormer is a transformer-based architecture that leverages a foundation model for robust feature extraction. It uniquely learns a comprehensive object representation by jointly encoding multiple template views and enriches these features with explicit 3D geometric priors using Normalized Object Coordinate Space (NOCS). A decoder then establishes robust 2D-3D correspondences to determine the final pose. Evaluated on the challenging BOP benchmarks, our integrated system demonstrates a strong balance between accuracy and efficiency, showcasing its practical applicability in both model-based and model-free scenarios.", "date": "2025-11-16", "categories": []}
{"id": "2511.11307", "title": "6D Strawberry Pose Estimation: Real-time and Edge AI Solutions Using Purely Synthetic Training Data", "abstract": "Automated and selective harvesting of fruits has become an important area of research, particularly due to challenges such as high costs and a shortage of seasonal labor in advanced economies. This paper focuses on 6D pose estimation of strawberries using purely synthetic data generated through a procedural pipeline for photorealistic rendering. We employ the YOLOX-6D-Pose algorithm, a single-shot approach that leverages the YOLOX backbone, known for its balance between speed and accuracy, and its support for edge inference. To address the lacking availability of training data, we introduce a robust and flexible pipeline for generating synthetic strawberry data from various 3D models via a procedural Blender pipeline, where we focus on enhancing the realism of the synthesized data in comparison to previous work to make it a valuable resource for training pose estimation algorithms. Quantitative evaluations indicate that our models achieve comparable accuracy on both the NVIDIA RTX 3090 and Jetson Orin Nano across several ADD-S metrics, with the RTX 3090 demonstrating superior processing speed. However, the Jetson Orin Nano is particularly suited for resource-constrained environments, making it an excellent choice for deployment in agricultural robotics. Qualitative assessments further confirm the model's performance, demonstrating its capability to accurately infer the poses of ripe and partially ripe strawberries, while facing challenges in detecting unripe specimens. This suggests opportunities for future improvements, especially in enhancing detection capabilities for unripe strawberries (if desired) by exploring variations in color. Furthermore, the methodology presented could be adapted easily for other fruits such as apples, peaches, and plums, thereby expanding its applicability and impact in the field of agricultural automation.", "date": "2025-11-14", "categories": []}
{"id": "2511.05996", "title": "Exploring Category-level Articulated Object Pose Tracking on SE(3) Manifolds", "abstract": "Articulated objects are prevalent in daily life and robotic manipulation tasks. However, compared to rigid objects, pose tracking for articulated objects remains an underexplored problem due to their inherent kinematic constraints. To address these challenges, this work proposes a novel point-pair-based pose tracking framework, termed \\textbf{PPF-Tracker}. The proposed framework first performs quasi-canonicalization of point clouds in the SE(3) Lie group space, and then models articulated objects using Point Pair Features (PPF) to predict pose voting parameters by leveraging the invariance properties of SE(3). Finally, semantic information of joint axes is incorporated to impose unified kinematic constraints across all parts of the articulated object. PPF-Tracker is systematically evaluated on both synthetic datasets and real-world scenarios, demonstrating strong generalization across diverse and challenging environments. Experimental results highlight the effectiveness and robustness of PPF-Tracker in multi-frame pose tracking of articulated objects. We believe this work can foster advances in robotics, embodied intelligence, and augmented reality. Codes are available at https://github.com/mengxh20/PPFTracker.", "date": "2025-11-08", "categories": []}
{"id": "2503.15106", "title": "Distilling 3D distinctive local descriptors for 6D pose estimation", "abstract": "Three-dimensional local descriptors are crucial for encoding geometric surface properties, making them essential for various point cloud understanding tasks. Among these descriptors, GeDi has demonstrated strong zero-shot 6D pose estimation capabilities but remains computationally impractical for real-world applications due to its expensive inference process. Can we retain GeDi's effectiveness while significantly improving its efficiency? In this paper, we explore this question by introducing a knowledge distillation framework that trains an efficient student model to regress local descriptors from a GeDi teacher. Our key contributions include: an efficient large-scale training procedure that ensures robustness to occlusions and partial observations while operating under compute and storage constraints, and a novel loss formulation that handles weak supervision from non-distinctive teacher descriptors. We validate our approach on five BOP Benchmark datasets and demonstrate a significant reduction in inference time while maintaining competitive performance with existing methods, bringing zero-shot 6D pose estimation closer to real-time feasibility. Project Website: https://tev-fbk.github.io/dGeDi/", "date": "2025-03-19", "categories": []}
{"id": "2511.02277", "title": "Are Euler angles a useful rotation parameterisation for pose estimation with Normalizing Flows?", "abstract": "Object pose estimation is a task that is of central importance in 3D Computer Vision. Given a target image and a canonical pose, a single point estimate may very often be sufficient; however, a probabilistic pose output is related to a number of benefits when pose is not unambiguous due to sensor and projection constraints or inherent object symmetries. With this paper, we explore the usefulness of using the well-known Euler angles parameterisation as a basis for a Normalizing Flows model for pose estimation. Isomorphic to spatial rotation, 3D pose has been parameterized in a number of ways, either in or out of the context of parameter estimation. We explore the idea that Euler angles, despite their shortcomings, may lead to useful models in a number of aspects, compared to a model built on a more complex parameterisation.", "date": "2025-11-04", "categories": []}
{"id": "2507.05698", "title": "Event-RGB Fusion for Spacecraft Pose Estimation Under Harsh Lighting", "abstract": "Spacecraft pose estimation is crucial for autonomous in-space operations, such as rendezvous, docking and on-orbit servicing. Vision-based pose estimation methods, which typically employ RGB imaging sensors, is a compelling solution for spacecraft pose estimation, but are challenged by harsh lighting conditions, which produce imaging artifacts such as glare, over-exposure, blooming and lens flare. Due to their much higher dynamic range, neuromorphic or event sensors are more resilient to extreme lighting conditions. However, event sensors generally have lower spatial resolution and suffer from reduced signal-to-noise ratio during periods of low relative motion. This work addresses these individual sensor limitations by introducing a sensor fusion approach combining RGB and event sensors. A beam-splitter prism was employed to achieve precise optical and temporal alignment. Then, a RANSAC-based technique was developed to fuse the information from the RGB and event channels to achieve pose estimation that leveraged the strengths of the two modalities. The pipeline was complemented by dropout uncertainty estimation to detect extreme conditions that affect either channel. To benchmark the performance of the proposed event-RGB fusion method, we collected a comprehensive real dataset of RGB and event data for satellite pose estimation in a laboratory setting under a variety of challenging illumination conditions. Encouraging results on the dataset demonstrate the efficacy of our event-RGB fusion approach and further supports the usage of event sensors for spacecraft pose estimation. To support community research on this topic, our dataset has been released publicly.", "date": "2025-07-08", "categories": []}
{"id": "2511.00120", "title": "VLM6D: VLM based 6Dof Pose Estimation based on RGB-D Images", "abstract": "The primary challenge in computer vision is precisely calculating the pose of 6D objects, however many current approaches are still fragile and have trouble generalizing from synthetic data to real-world situations with fluctuating lighting, textureless objects, and significant occlusions. To address these limitations, VLM6D, a novel dual-stream architecture that leverages the distinct strengths of visual and geometric data from RGB-D input for robust and precise pose estimation. Our framework uniquely integrates two specialized encoders: a powerful, self-supervised Vision Transformer (DINOv2) processes the RGB modality, harnessing its rich, pre-trained understanding of visual grammar to achieve remarkable resilience against texture and lighting variations. Concurrently, a PointNet++ encoder processes the 3D point cloud derived from depth data, enabling robust geometric reasoning that excels even with the sparse, fragmented data typical of severe occlusion. These complementary feature streams are effectively fused to inform a multi task prediction head. We demonstrate through comprehensive experiments that VLM6D obtained new SOTA performance on the challenging Occluded-LineMOD, validating its superior robustness and accuracy.", "date": "2025-10-31", "categories": []}
{"id": "2511.01501", "title": "SE(3)-PoseFlow: Estimating 6D Pose Distributions for Uncertainty-Aware Robotic Manipulation", "abstract": "Object pose estimation is a fundamental problem in robotics and computer vision, yet it remains challenging due to partial observability, occlusions, and object symmetries, which inevitably lead to pose ambiguity and multiple hypotheses consistent with the same observation. While deterministic deep networks achieve impressive performance under well-constrained conditions, they are often overconfident and fail to capture the multi-modality of the underlying pose distribution. To address these challenges, we propose a novel probabilistic framework that leverages flow matching on the SE(3) manifold for estimating 6D object pose distributions. Unlike existing methods that regress a single deterministic output, our approach models the full pose distribution with a sample-based estimate and enables reasoning about uncertainty in ambiguous cases such as symmetric objects or severe occlusions. We achieve state-of-the-art results on Real275, YCB-V, and LM-O, and demonstrate how our sample-based pose estimates can be leveraged in downstream robotic manipulation tasks such as active perception for disambiguating uncertain viewpoints or guiding grasp synthesis in an uncertainty-aware manner.", "date": "2025-11-03", "categories": []}
{"id": "2510.18521", "title": "RayPose: Ray Bundling Diffusion for Template Views in Unseen 6D Object Pose Estimation", "abstract": "Typical template-based object pose pipelines estimate the pose by retrieving the closest matching template and aligning it with the observed image. However, failure to retrieve the correct template often leads to inaccurate pose predictions. To address this, we reformulate template-based object pose estimation as a ray alignment problem, where the viewing directions from multiple posed template images are learned to align with a non-posed query image. Inspired by recent progress in diffusion-based camera pose estimation, we embed this formulation into a diffusion transformer architecture that aligns a query image with a set of posed templates. We reparameterize object rotation using object-centered camera rays and model object translation by extending scale-invariant translation estimation to dense translation offsets. Our model leverages geometric priors from the templates to guide accurate query pose inference. A coarse-to-fine training strategy based on narrowed template sampling improves performance without modifying the network architecture. Extensive experiments across multiple benchmark datasets show competitive results of our method compared to state-of-the-art approaches in unseen object pose estimation.", "date": "2025-10-21", "categories": []}
{"id": "2505.16144", "title": "GMatch: A Lightweight, Geometry-Constrained Keypoint Matcher for Zero-Shot 6DoF Pose Estimation in Robotic Grasp Tasks", "abstract": "6DoF object pose estimation is fundamental to robotic grasp tasks. While recent learning-based methods achieve high accuracy, their computational demands hinder deployment on resource-constrained mobile platforms. In this work, we revisit the classical keypoint matching paradigm and propose GMatch, a lightweight, geometry-constrained keypoint matcher that can run efficiently on embedded CPU-only platforms. GMatch works with keypoint descriptors and it uses a set of geometric constraints to establishes inherent ambiguities between features extracted by descriptors, thus giving a globally consistent correspondences from which 6DoF pose can be easily solved. We benchmark GMatch on the HOPE and YCB-Video datasets, where our method beats existing keypoint matchers (both feature-based and geometry-based) among three commonly used descriptors and approaches the SOTA zero-shot method on texture-rich objects with much more humble devices. The method is further deployed on a LoCoBot mobile manipulator, enabling a one-shot grasp pipeline that demonstrates high task success rates in real-world experiments. In a word, by its lightweight and white-box nature, GMatch offers a practical solution for resource-limited robotic systems, and although currently bottlenecked by descriptor quality, the framework presents a promising direction towards robust yet efficient pose estimation.\n  Code will be released soon under Mozilla Public License.", "date": "2025-05-22", "categories": []}
{"id": "2510.16777", "title": "GS2POSE: Marry Gaussian Splatting to 6D Object Pose Estimation", "abstract": "Accurate 6D pose estimation of 3D objects is a fundamental task in computer vision, and current research typically predicts the 6D pose by establishing correspondences between 2D image features and 3D model features. However, these methods often face difficulties with textureless objects and varying illumination conditions. To overcome these limitations, we propose GS2POSE, a novel approach for 6D object pose estimation. GS2POSE formulates a pose regression algorithm inspired by the principles of Bundle Adjustment (BA). By leveraging Lie algebra, we extend the capabilities of 3DGS to develop a pose-differentiable rendering pipeline, which iteratively optimizes the pose by comparing the input image to the rendered image. Additionally, GS2POSE updates color parameters within the 3DGS model, enhancing its adaptability to changes in illumination. Compared to previous models, GS2POSE demonstrates accuracy improvements of 1.4\\%, 2.8\\% and 2.5\\% on the T-LESS, LineMod-Occlusion and LineMod datasets, respectively.", "date": "2025-10-19", "categories": []}
{"id": "2407.12207", "title": "NeuSurfEmb: A Complete Pipeline for Dense Correspondence-based 6D Object Pose Estimation without CAD Models", "abstract": "State-of-the-art approaches for 6D object pose estimation assume the availability of CAD models and require the user to manually set up physically-based rendering (PBR) pipelines for synthetic training data generation. Both factors limit the application of these methods in real-world scenarios. In this work, we present a pipeline that does not require CAD models and allows training a state-of-the-art pose estimator requiring only a small set of real images as input. Our method is based on a NeuS2 object representation, that we learn through a semi-automated procedure based on Structure-from-Motion (SfM) and object-agnostic segmentation. We exploit the novel-view synthesis ability of NeuS2 and simple cut-and-paste augmentation to automatically generate photorealistic object renderings, which we use to train the correspondence-based SurfEmb pose estimator. We evaluate our method on the LINEMOD-Occlusion dataset, extensively studying the impact of its individual components and showing competitive performance with respect to approaches based on CAD models and PBR data. We additionally demonstrate the ease of use and effectiveness of our pipeline on self-collected real-world objects, showing that our method outperforms state-of-the-art CAD-model-free approaches, with better accuracy and robustness to mild occlusions. To allow the robotics community to benefit from this system, we will publicly release it at https://www.github.com/ethz-asl/neusurfemb.", "date": "2024-07-16", "categories": []}
{"id": "2510.10177", "title": "HccePose(BF): Predicting Front & Back Surfaces to Construct Ultra-Dense 2D-3D Correspondences for Pose Estimation", "abstract": "In pose estimation for seen objects, a prevalent pipeline involves using neural networks to predict dense 3D coordinates of the object surface on 2D images, which are then used to establish dense 2D-3D correspondences. However, current methods primarily focus on more efficient encoding techniques to improve the precision of predicted 3D coordinates on the object's front surface, overlooking the potential benefits of incorporating the back surface and interior of the object. To better utilize the full surface and interior of the object, this study predicts 3D coordinates of both the object's front and back surfaces and densely samples 3D coordinates between them. This process creates ultra-dense 2D-3D correspondences, effectively enhancing pose estimation accuracy based on the Perspective-n-Point (PnP) algorithm. Additionally, we propose Hierarchical Continuous Coordinate Encoding (HCCE) to provide a more accurate and efficient representation of front and back surface coordinates. Experimental results show that, compared to existing state-of-the-art (SOTA) methods on the BOP website, the proposed approach outperforms across seven classic BOP core datasets. Code is available at https://github.com/WangYuLin-SEU/HCCEPose.", "date": "2025-10-11", "categories": []}
{"id": "2510.10434", "title": "MonoSE(3)-Diffusion: A Monocular SE(3) Diffusion Framework for Robust Camera-to-Robot Pose Estimation", "abstract": "We propose MonoSE(3)-Diffusion, a monocular SE(3) diffusion framework that formulates markerless, image-based robot pose estimation as a conditional denoising diffusion process. The framework consists of two processes: a visibility-constrained diffusion process for diverse pose augmentation and a timestep-aware reverse process for progressive pose refinement. The diffusion process progressively perturbs ground-truth poses to noisy transformations for training a pose denoising network. Importantly, we integrate visibility constraints into the process, ensuring the transformations remain within the camera field of view. Compared to the fixed-scale perturbations used in current methods, the diffusion process generates in-view and diverse training poses, thereby improving the network generalization capability. Furthermore, the reverse process iteratively predicts the poses by the denoising network and refines pose estimates by sampling from the diffusion posterior of current timestep, following a scheduled coarse-to-fine procedure. Moreover, the timestep indicates the transformation scales, which guide the denoising network to achieve more accurate pose predictions. The reverse process demonstrates higher robustness than direct prediction, benefiting from its timestep-aware refinement scheme. Our approach demonstrates improvements across two benchmarks (DREAM and RoboKeyGen), achieving a notable AUC of 66.75 on the most challenging dataset, representing a 32.3% gain over the state-of-the-art.", "date": "2025-10-12", "categories": []}
{"id": "2510.10933", "title": "DKPMV: Dense Keypoints Fusion from Multi-View RGB Frames for 6D Pose Estimation of Textureless Objects", "abstract": "6D pose estimation of textureless objects is valuable for industrial robotic applications, yet remains challenging due to the frequent loss of depth information. Current multi-view methods either rely on depth data or insufficiently exploit multi-view geometric cues, limiting their performance. In this paper, we propose DKPMV, a pipeline that achieves dense keypoint-level fusion using only multi-view RGB images as input. We design a three-stage progressive pose optimization strategy that leverages dense multi-view keypoint geometry information. To enable effective dense keypoint fusion, we enhance the keypoint network with attentional aggregation and symmetry-aware training, improving prediction accuracy and resolving ambiguities on symmetric objects. Extensive experiments on the ROBI dataset demonstrate that DKPMV outperforms state-of-the-art multi-view RGB approaches and even surpasses the RGB-D methods in the majority of cases. The code will be available soon.", "date": "2025-10-13", "categories": []}
{"id": "2510.11687", "title": "Beyond 'Templates': Category-Agnostic Object Pose, Size, and Shape Estimation from a Single View", "abstract": "Estimating an object's 6D pose, size, and shape from visual input is a fundamental problem in computer vision, with critical applications in robotic grasping and manipulation. Existing methods either rely on object-specific priors such as CAD models or templates, or suffer from limited generalization across categories due to pose-shape entanglement and multi-stage pipelines. In this work, we propose a unified, category-agnostic framework that simultaneously predicts 6D pose, size, and dense shape from a single RGB-D image, without requiring templates, CAD models, or category labels at test time. Our model fuses dense 2D features from vision foundation models with partial 3D point clouds using a Transformer encoder enhanced by a Mixture-of-Experts, and employs parallel decoders for pose-size estimation and shape reconstruction, achieving real-time inference at 28 FPS. Trained solely on synthetic data from 149 categories in the SOPE dataset, our framework is evaluated on four diverse benchmarks SOPE, ROPE, ObjaversePose, and HANDAL, spanning over 300 categories. It achieves state-of-the-art accuracy on seen categories while demonstrating remarkably strong zero-shot generalization to unseen real-world objects, establishing a new standard for open-set 6D understanding in robotics and embodied AI.", "date": "2025-10-13", "categories": []}
{"id": "2408.13147", "title": "ShapeICP: Iterative Category-level Object Pose and Shape Estimation from Depth", "abstract": "Category-level object pose and shape estimation from a single depth image has recently drawn research attention due to its potential utility for tasks such as robotics manipulation. The task is particularly challenging because the three unknowns, object pose, object shape, and model-to-measurement correspondences, are compounded together, but only a single view of depth measurements is provided. Most of the prior work heavily relies on data-driven approaches to obtain solutions to at least one of the unknowns, and typically two, risking generalization failures if not designed and trained carefully. The shape representations used in the prior work also mainly focus on point clouds and signed distance fields (SDFs). In stark contrast to the prior work, we approach the problem using an iterative estimation method that does not require learning from pose-annotated data. Moreover, we construct and adopt a novel mesh-based object active shape model (ASM), which additionally maintains vertex connectivity compared to the commonly used point-based object ASM. Our algorithm, ShapeICP, is based on the iterative closest point (ICP) algorithm but is equipped with additional features for the category-level pose and shape estimation task. Although not using pose-annotated data, ShapeICP surpasses many data-driven approaches that rely on pose data for training, opening up a new solution space for researchers to consider.", "date": "2024-08-23", "categories": []}
{"id": "2510.04125", "title": "Joint Learning of Pose Regression and Denoising Diffusion with Score Scaling Sampling for Category-level 6D Pose Estimation", "abstract": "Latest diffusion models have shown promising results in category-level 6D object pose estimation by modeling the conditional pose distribution with depth image input. The existing methods, however, suffer from slow convergence during training, learning its encoder with the diffusion denoising network in end-to-end fashion, and require an additional network that evaluates sampled pose hypotheses to filter out low-quality pose candidates. In this paper, we propose a novel pipeline that tackles these limitations by two key components. First, the proposed method pretrains the encoder with the direct pose regression head, and jointly learns the networks via the regression head and the denoising diffusion head, significantly accelerating training convergence while achieving higher accuracy. Second, sampling guidance via time-dependent score scaling is proposed s.t. the exploration-exploitation trade-off is effectively taken, eliminating the need for the additional evaluation network. The sampling guidance maintains multi-modal characteristics of symmetric objects at early denoising steps while ensuring high-quality pose generation at final steps. Extensive experiments on multiple benchmarks including REAL275, HouseCat6D, and ROPE, demonstrate that the proposed method, simple yet effective, achieves state-of-the-art accuracies even with single-pose inference, while being more efficient in both training and inference.", "date": "2025-10-05", "categories": []}
{"id": "2504.07955", "title": "BoxDreamer: Dreaming Box Corners for Generalizable Object Pose Estimation", "abstract": "This paper presents a generalizable RGB-based approach for object pose estimation, specifically designed to address challenges in sparse-view settings. While existing methods can estimate the poses of unseen objects, their generalization ability remains limited in scenarios involving occlusions and sparse reference views, restricting their real-world applicability. To overcome these limitations, we introduce corner points of the object bounding box as an intermediate representation of the object pose. The 3D object corners can be reliably recovered from sparse input views, while the 2D corner points in the target view are estimated through a novel reference-based point synthesizer, which works well even in scenarios involving occlusions. As object semantic points, object corners naturally establish 2D-3D correspondences for object pose estimation with a PnP algorithm. Extensive experiments on the YCB-Video and Occluded-LINEMOD datasets show that our approach outperforms state-of-the-art methods, highlighting the effectiveness of the proposed representation and significantly enhancing the generalization capabilities of object pose estimation, which is crucial for real-world applications.", "date": "2025-04-10", "categories": []}
{"id": "2509.23647", "title": "Color-Pair Guided Robust Zero-Shot 6D Pose Estimation and Tracking of Cluttered Objects on Edge Devices", "abstract": "Robust 6D pose estimation of novel objects under challenging illumination remains a significant challenge, often requiring a trade-off between accurate initial pose estimation and efficient real-time tracking. We present a unified framework explicitly designed for efficient execution on edge devices, which synergizes a robust initial estimation module with a fast motion-based tracker. The key to our approach is a shared, lighting-invariant color-pair feature representation that forms a consistent foundation for both stages. For initial estimation, this feature facilitates robust registration between the live RGB-D view and the object's 3D mesh. For tracking, the same feature logic validates temporal correspondences, enabling a lightweight model to reliably regress the object's motion. Extensive experiments on benchmark datasets demonstrate that our integrated approach is both effective and robust, providing competitive pose estimation accuracy while maintaining high-fidelity tracking even through abrupt pose changes.", "date": "2025-09-28", "categories": []}
{"id": "2509.24572", "title": "SCOPE: Semantic Conditioning for Sim2Real Category-Level Object Pose Estimation in Robotics", "abstract": "Object manipulation requires accurate object pose estimation. In open environments, robots encounter unknown objects, which requires semantic understanding in order to generalize both to known categories and beyond. To resolve this challenge, we present SCOPE, a diffusion-based category-level object pose estimation model that eliminates the need for discrete category labels by leveraging DINOv2 features as continuous semantic priors. By combining these DINOv2 features with photorealistic training data and a noise model for point normals, we reduce the Sim2Real gap in category-level object pose estimation. Furthermore, injecting the continuous semantic priors via cross-attention enables SCOPE to learn canonicalized object coordinate systems across object instances beyond the distribution of known categories. SCOPE outperforms the current state of the art in synthetically trained category-level object pose estimation, achieving a relative improvement of 31.9\\% on the 5$^\\circ$5cm metric. Additional experiments on two instance-level datasets demonstrate generalization beyond known object categories, enabling grasping of unseen objects from unknown categories with a success rate of up to 100\\%. Code available: https://github.com/hoenigpeter/scope.", "date": "2025-09-29", "categories": []}
{"id": "2509.21927", "title": "SingRef6D: Monocular Novel Object Pose Estimation with a Single RGB Reference", "abstract": "Recent 6D pose estimation methods demonstrate notable performance but still face some practical limitations. For instance, many of them rely heavily on sensor depth, which may fail with challenging surface conditions, such as transparent or highly reflective materials. In the meantime, RGB-based solutions provide less robust matching performance in low-light and texture-less scenes due to the lack of geometry information. Motivated by these, we propose SingRef6D, a lightweight pipeline requiring only a single RGB image as a reference, eliminating the need for costly depth sensors, multi-view image acquisition, or training view synthesis models and neural fields. This enables SingRef6D to remain robust and capable even under resource-limited settings where depth or dense templates are unavailable. Our framework incorporates two key innovations. First, we propose a token-scaler-based fine-tuning mechanism with a novel optimization loss on top of Depth-Anything v2 to enhance its ability to predict accurate depth, even for challenging surfaces. Our results show a 14.41% improvement (in $\\delta_{1.05}$) on REAL275 depth prediction compared to Depth-Anything v2 (with fine-tuned head). Second, benefiting from depth availability, we introduce a depth-aware matching process that effectively integrates spatial relationships within LoFTR, enabling our system to handle matching for challenging materials and lighting conditions. Evaluations of pose estimation on the REAL275, ClearPose, and Toyota-Light datasets show that our approach surpasses state-of-the-art methods, achieving a 6.1% improvement in average recall.", "date": "2025-09-26", "categories": []}
{"id": "2406.11743", "title": "Domain Generalization for In-Orbit 6D Pose Estimation", "abstract": "We address the problem of estimating the relative 6D pose, i.e., position and orientation, of a target spacecraft, from a monocular image, a key capability for future autonomous Rendezvous and Proximity Operations. Due to the difficulty of acquiring large sets of real images, spacecraft pose estimation networks are exclusively trained on synthetic ones. However, because those images do not capture the illumination conditions encountered in orbit, pose estimation networks face a domain gap problem, i.e., they do not generalize to real images. Our work introduces a method that bridges this domain gap. It relies on a novel, end-to-end, neural-based architecture as well as a novel learning strategy. This strategy improves the domain generalization abilities of the network through multi-task learning and aggressive data augmentation policies, thereby enforcing the network to learn domain-invariant features. We demonstrate that our method effectively closes the domain gap, achieving state-of-the-art accuracy on the widespread SPEED+ dataset. Finally, ablation studies assess the impact of key components of our method on its generalization abilities.", "date": "2024-06-17", "categories": []}
{"id": "2509.13591", "title": "Object Pose Estimation through Dexterous Touch", "abstract": "Robust object pose estimation is essential for manipulation and interaction tasks in robotics, particularly in scenarios where visual data is limited or sensitive to lighting, occlusions, and appearances. Tactile sensors often offer limited and local contact information, making it challenging to reconstruct the pose from partial data. Our approach uses sensorimotor exploration to actively control a robot hand to interact with the object. We train with Reinforcement Learning (RL) to explore and collect tactile data. The collected 3D point clouds are used to iteratively refine the object's shape and pose. In our setup, one hand holds the object steady while the other performs active exploration. We show that our method can actively explore an object's surface to identify critical pose features without prior knowledge of the object's geometry. Supplementary material and more demonstrations will be provided at https://amirshahid.github.io/BimanualTactilePose .", "date": "2025-09-16", "categories": []}
{"id": "2509.11680", "title": "IMD: A 6-DoF Pose Estimation Benchmark for Industrial Metallic Objects", "abstract": "Object 6DoF (6D) pose estimation is essential for robotic perception, especially in industrial settings. It enables robots to interact with the environment and manipulate objects. However, existing benchmarks on object 6D pose estimation primarily use everyday objects with rich textures and low-reflectivity, limiting model generalization to industrial scenarios where objects are often metallic, texture-less, and highly reflective. To address this gap, we propose a novel dataset and benchmark namely \\textit{Industrial Metallic Dataset (IMD)}, tailored for industrial applications. Our dataset comprises 45 true-to-scale industrial components, captured with an RGB-D camera under natural indoor lighting and varied object arrangements to replicate real-world conditions. The benchmark supports three tasks, including video object segmentation, 6D pose tracking, and one-shot 6D pose estimation. We evaluate existing state-of-the-art models, including XMem and SAM2 for segmentation, and BundleTrack and BundleSDF for pose estimation, to assess model performance in industrial contexts. Evaluation results show that our industrial dataset is more challenging than existing household object datasets. This benchmark provides the baseline for developing and comparing segmentation and pose estimation algorithms that better generalize to industrial robotics scenarios.", "date": "2025-09-15", "categories": []}
{"id": "2502.17434", "title": "V-HOP: Visuo-Haptic 6D Object Pose Tracking", "abstract": "Humans naturally integrate vision and haptics for robust object perception during manipulation. The loss of either modality significantly degrades performance. Inspired by this multisensory integration, prior object pose estimation research has attempted to combine visual and haptic/tactile feedback. Although these works demonstrate improvements in controlled environments or synthetic datasets, they often underperform vision-only approaches in real-world settings due to poor generalization across diverse grippers, sensor layouts, or sim-to-real environments. Furthermore, they typically estimate the object pose for each frame independently, resulting in less coherent tracking over sequences in real-world deployments. To address these limitations, we introduce a novel unified haptic representation that effectively handles multiple gripper embodiments. Building on this representation, we introduce a new visuo-haptic transformer-based object pose tracker that seamlessly integrates visual and haptic input. We validate our framework in our dataset and the Feelsight dataset, demonstrating significant performance improvement on challenging sequences. Notably, our method achieves superior generalization and robustness across novel embodiments, objects, and sensor types (both taxel-based and vision-based tactile sensors). In real-world experiments, we demonstrate that our approach outperforms state-of-the-art visual trackers by a large margin. We further show that we can achieve precise manipulation tasks by incorporating our real-time object tracking result into motion plans, underscoring the advantages of visuo-haptic perception. Project website: https://ivl.cs.brown.edu/research/v-hop", "date": "2025-02-24", "categories": []}
{"id": "2509.07978", "title": "One View, Many Worlds: Single-Image to 3D Object Meets Generative Domain Randomization for One-Shot 6D Pose Estimation", "abstract": "Estimating the 6D pose of arbitrary unseen objects from a single reference image is critical for robotics operating in the long-tail of real-world instances. However, this setting is notoriously challenging: 3D models are rarely available, single-view reconstructions lack metric scale, and domain gaps between generated models and real-world images undermine robustness. We propose OnePoseViaGen, a pipeline that tackles these challenges through two key components. First, a coarse-to-fine alignment module jointly refines scale and pose by combining multi-view feature matching with render-and-compare refinement. Second, a text-guided generative domain randomization strategy diversifies textures, enabling effective fine-tuning of pose estimators with synthetic data. Together, these steps allow high-fidelity single-view 3D generation to support reliable one-shot 6D pose estimation. On challenging benchmarks (YCBInEOAT, Toyota-Light, LM-O), OnePoseViaGen achieves state-of-the-art performance far surpassing prior approaches. We further demonstrate robust dexterous grasping with a real robot hand, validating the practicality of our method in real-world manipulation. Project page: https://gzwsama.github.io/OnePoseviaGen.github.io/", "date": "2025-09-09", "categories": []}
{"id": "2509.06000", "title": "Motion Aware ViT-based Framework for Monocular 6-DoF Spacecraft Pose Estimation", "abstract": "Monocular 6-DoF pose estimation plays an important role in multiple spacecraft missions. Most existing pose estimation approaches rely on single images with static keypoint localisation, failing to exploit valuable temporal information inherent to space operations. In this work, we adapt a deep learning framework from human pose estimation to the spacecraft pose estimation domain that integrates motion-aware heatmaps and optical flow to capture motion dynamics. Our approach combines image features from a Vision Transformer (ViT) encoder with motion cues from a pre-trained optical flow model to localise 2D keypoints. Using the estimates, a Perspective-n-Point (PnP) solver recovers 6-DoF poses from known 2D-3D correspondences. We train and evaluate our method on the SPADES-RGB dataset and further assess its generalisation on real and synthetic data from the SPARK-2024 dataset. Overall, our approach demonstrates improved performance over single-image baselines in both 2D keypoint localisation and 6-DoF pose estimation. Furthermore, it shows promising generalisation capabilities when testing on different data distributions.", "date": "2025-09-07", "categories": []}
{"id": "2406.18453", "title": "Towards Human-Level 3D Relative Pose Estimation: Generalizable, Training-Free, with Single Reference", "abstract": "Humans can easily deduce the relative pose of a previously unseen object, without labeling or training, given only a single query-reference image pair. This is arguably achieved by incorporating i) 3D/2.5D shape perception from a single image, ii) render-and-compare simulation, and iii) rich semantic cue awareness to furnish (coarse) reference-query correspondence. Motivated by this, we propose a novel 3D generalizable relative pose estimation method by elaborating 3D/2.5D shape perception with a 2.5D shape from an RGB-D reference, fulfilling the render-and-compare paradigm with an off-the-shelf differentiable renderer, and leveraging the semantic cues from a pretrained model like DINOv2. Specifically, our differentiable renderer takes the 2.5D rotatable mesh textured by the RGB and the semantic maps (obtained by DINOv2 from the RGB input), then renders new RGB and semantic maps (with back-surface culling) under a novel rotated view. The refinement loss comes from comparing the rendered RGB and semantic maps with the query ones, back-propagating the gradients through the differentiable renderer to refine the 3D relative pose. As a result, \\emph{our method can be readily applied to unseen objects, given only a single RGB-D reference, without labeling or training}. Extensive experiments on LineMOD, LM-O, and YCB-V show that our training-free method significantly outperforms the state-of-the-art supervised methods, especially under the rigorous \\texttt{Acc@5/10/15}$^\\circ$ metrics and the challenging cross-dataset settings.", "date": "2024-06-26", "categories": []}
{"id": "2504.02617", "title": "PicoPose: Progressive Pixel-to-Pixel Correspondence Learning for Novel Object Pose Estimation", "abstract": "RGB-based novel object pose estimation is critical for rapid deployment in robotic applications, yet zero-shot generalization remains a key challenge. In this paper, we introduce PicoPose, a novel framework designed to tackle this task using a three-stage pixel-to-pixel correspondence learning process. Firstly, PicoPose matches features from the RGB observation with those from rendered object templates, identifying the best-matched template and establishing coarse correspondences. Secondly, PicoPose smooths the correspondences by globally regressing a 2D affine transformation, including in-plane rotation, scale, and 2D translation, from the coarse correspondence map. Thirdly, PicoPose applies the affine transformation to the feature map of the best-matched template and learns correspondence offsets within local regions to achieve fine-grained correspondences. By progressively refining the correspondences, PicoPose significantly improves the accuracy of object poses computed via PnP/RANSAC. PicoPose achieves state-of-the-art performance on the seven core datasets of the BOP benchmark, demonstrating exceptional generalization to novel objects. Code and trained models are available at https://github.com/foollh/PicoPose.", "date": "2025-04-03", "categories": []}
{"id": "2508.17746", "title": "DroneKey: Drone 3D Pose Estimation in Image Sequences using Gated Key-representation and Pose-adaptive Learning", "abstract": "Estimating the 3D pose of a drone is important for anti-drone systems, but existing methods struggle with the unique challenges of drone keypoint detection. Drone propellers serve as keypoints but are difficult to detect due to their high visual similarity and diversity of poses. To address these challenges, we propose DroneKey, a framework that combines a 2D keypoint detector and a 3D pose estimator specifically designed for drones. In the keypoint detection stage, we extract two key-representations (intermediate and compact) from each transformer encoder layer and optimally combine them using a gated sum. We also introduce a pose-adaptive Mahalanobis distance in the loss function to ensure stable keypoint predictions across extreme poses. We built new datasets of drone 2D keypoints and 3D pose to train and evaluate our method, which have been publicly released. Experiments show that our method achieves an AP of 99.68% (OKS) in keypoint detection, outperforming existing methods. Ablation studies confirm that the pose-adaptive Mahalanobis loss function improves keypoint prediction stability and accuracy. Additionally, improvements in the encoder design enable real-time processing at 44 FPS. For 3D pose estimation, our method achieved an MAE-angle of 10.62{\\deg}, an RMSE of 0.221m, and an MAE-absolute of 0.076m, demonstrating high accuracy and reliability. The code and dataset are available at https://github.com/kkanuseobin/DroneKey.", "date": "2025-08-25", "categories": []}
{"id": "2508.15972", "title": "UnPose: Uncertainty-Guided Diffusion Priors for Zero-Shot Pose Estimation", "abstract": "Estimating the 6D pose of novel objects is a fundamental yet challenging problem in robotics, often relying on access to object CAD models. However, acquiring such models can be costly and impractical. Recent approaches aim to bypass this requirement by leveraging strong priors from foundation models to reconstruct objects from single or multi-view images, but typically require additional training or produce hallucinated geometry. To this end, we propose UnPose, a novel framework for zero-shot, model-free 6D object pose estimation and reconstruction that exploits 3D priors and uncertainty estimates from a pre-trained diffusion model. Specifically, starting from a single-view RGB-D frame, UnPose uses a multi-view diffusion model to estimate an initial 3D model using 3D Gaussian Splatting (3DGS) representation, along with pixel-wise epistemic uncertainty estimates. As additional observations become available, we incrementally refine the 3DGS model by fusing new views guided by the diffusion model's uncertainty, thereby continuously improving the pose estimation accuracy and 3D reconstruction quality. To ensure global consistency, the diffusion prior-generated views and subsequent observations are further integrated in a pose graph and jointly optimized into a coherent 3DGS field. Extensive experiments demonstrate that UnPose significantly outperforms existing approaches in both 6D pose estimation accuracy and 3D reconstruction quality. We further showcase its practical applicability in real-world robotic manipulation tasks.", "date": "2025-08-21", "categories": []}
{"id": "2508.14358", "title": "Learning Point Cloud Representations with Pose Continuity for Depth-Based Category-Level 6D Object Pose Estimation", "abstract": "Category-level object pose estimation aims to predict the 6D pose and 3D size of objects within given categories. Existing approaches for this task rely solely on 6D poses as supervisory signals without explicitly capturing the intrinsic continuity of poses, leading to inconsistencies in predictions and reduced generalization to unseen poses. To address this limitation, we propose HRC-Pose, a novel depth-only framework for category-level object pose estimation, which leverages contrastive learning to learn point cloud representations that preserve the continuity of 6D poses. HRC-Pose decouples object pose into rotation and translation components, which are separately encoded and leveraged throughout the network. Specifically, we introduce a contrastive learning strategy for multi-task, multi-category scenarios based on our 6D pose-aware hierarchical ranking scheme, which contrasts point clouds from multiple categories by considering rotational and translational differences as well as categorical information. We further design pose estimation modules that separately process the learned rotation-aware and translation-aware embeddings. Our experiments demonstrate that HRC-Pose successfully learns continuous feature spaces. Results on REAL275 and CAMERA25 benchmarks show that our method consistently outperforms existing depth-only state-of-the-art methods and runs in real-time, demonstrating its effectiveness and potential for real-world applications. Our code is at https://github.com/zhujunli1993/HRC-Pose.", "date": "2025-08-20", "categories": []}
{"id": "2508.14767", "title": "Fusing Monocular RGB Images with AIS Data to Create a 6D Pose Estimation Dataset for Marine Vessels", "abstract": "The paper presents a novel technique for creating a 6D pose estimation dataset for marine vessels by fusing monocular RGB images with Automatic Identification System (AIS) data. The proposed technique addresses the limitations of relying purely on AIS for location information, caused by issues like equipment reliability, data manipulation, and transmission delays. By combining vessel detections from monocular RGB images, obtained using an object detection network (YOLOX-X), with AIS messages, the technique generates 3D bounding boxes that represent the vessels' 6D poses, i.e. spatial and rotational dimensions. The paper evaluates different object detection models to locate vessels in image space. We also compare two transformation methods (homography and Perspective-n-Point) for aligning AIS data with image coordinates. The results of our work demonstrate that the Perspective-n-Point (PnP) method achieves a significantly lower projection error compared to homography-based approaches used before, and the YOLOX-X model achieves a mean Average Precision (mAP) of 0.80 at an Intersection over Union (IoU) threshold of 0.5 for relevant vessel classes. We show indication that our approach allows the creation of a 6D pose estimation dataset without needing manual annotation. Additionally, we introduce the Boats on Nordelbe Kehrwieder (BONK-pose), a publicly available dataset comprising 3753 images with 3D bounding box annotations for pose estimation, created by our data fusion approach. This dataset can be used for training and evaluating 6D pose estimation networks. In addition we introduce a set of 1000 images with 2D bounding box annotations for ship detection from the same scene.", "date": "2025-08-20", "categories": []}
{"id": "2508.13623", "title": "RCGNet: RGB-based Category-Level 6D Object Pose Estimation with Geometric Guidance", "abstract": "While most current RGB-D-based category-level object pose estimation methods achieve strong performance, they face significant challenges in scenes lacking depth information. In this paper, we propose a novel category-level object pose estimation approach that relies solely on RGB images. This method enables accurate pose estimation in real-world scenarios without the need for depth data. Specifically, we design a transformer-based neural network for category-level object pose estimation, where the transformer is employed to predict and fuse the geometric features of the target object. To ensure that these predicted geometric features faithfully capture the object's geometry, we introduce a geometric feature-guided algorithm, which enhances the network's ability to effectively represent the object's geometric information. Finally, we utilize the RANSAC-PnP algorithm to compute the object's pose, addressing the challenges associated with variable object scales in pose estimation. Experimental results on benchmark datasets demonstrate that our approach is not only highly efficient but also achieves superior accuracy compared to previous RGB-based methods. These promising results offer a new perspective for advancing category-level object pose estimation using RGB images.", "date": "2025-08-19", "categories": []}
{"id": "2508.13775", "title": "MR6D: Benchmarking 6D Pose Estimation for Mobile Robots", "abstract": "Existing 6D pose estimation datasets primarily focus on small household objects typically handled by robot arm manipulators, limiting their relevance to mobile robotics. Mobile platforms often operate without manipulators, interact with larger objects, and face challenges such as long-range perception, heavy self-occlusion, and diverse camera perspectives. While recent models generalize well to unseen objects, evaluations remain confined to household-like settings that overlook these factors. We introduce MR6D, a dataset designed for 6D pose estimation for mobile robots in industrial environments. It includes 92 real-world scenes featuring 16 unique objects across static and dynamic interactions. MR6D captures the challenges specific to mobile platforms, including distant viewpoints, varied object configurations, larger object sizes, and complex occlusion/self-occlusion patterns. Initial experiments reveal that current 6D pipelines underperform in these settings, with 2D segmentation being another hurdle. MR6D establishes a foundation for developing and evaluating pose estimation methods tailored to the demands of mobile robotics. The dataset is available at https://huggingface.co/datasets/anas-gouda/mr6d.", "date": "2025-08-19", "categories": []}
{"id": "2504.07335", "title": "DLTPose: 6DoF Pose Estimation From Accurate Dense Surface Point Estimates", "abstract": "We propose DLTPose, a novel method for 6DoF object pose estimation from RGBD images that combines the accuracy of sparse keypoint methods with the robustness of dense pixel-wise predictions. DLTPose predicts per-pixel radial distances to a set of minimally four keypoints, which are then fed into our novel Direct Linear Transform (DLT) formulation to produce accurate 3D object frame surface estimates, leading to better 6DoF pose estimation. Additionally, we introduce a novel symmetry-aware keypoint ordering approach, designed to handle object symmetries that otherwise cause inconsistencies in keypoint assignments. Previous keypoint-based methods relied on fixed keypoint orderings, which failed to account for the multiple valid configurations exhibited by symmetric objects, which our ordering approach exploits to enhance the model's ability to learn stable keypoint representations. Extensive experiments on the benchmark LINEMOD, Occlusion LINEMOD and YCB-Video datasets show that DLTPose outperforms existing methods, especially for symmetric and occluded objects. The code is available at https://anonymous.4open.science/r/DLTPose_/ .", "date": "2025-04-09", "categories": []}
{"id": "2508.11950", "title": "DynamicPose: Real-time and Robust 6D Object Pose Tracking for Fast-Moving Cameras and Objects", "abstract": "We present DynamicPose, a retraining-free 6D pose tracking framework that improves tracking robustness in fast-moving camera and object scenarios. Previous work is mainly applicable to static or quasi-static scenes, and its performance significantly deteriorates when both the object and the camera move rapidly. To overcome these challenges, we propose three synergistic components: (1) A visual-inertial odometry compensates for the shift in the Region of Interest (ROI) caused by camera motion; (2) A depth-informed 2D tracker corrects ROI deviations caused by large object translation; (3) A VIO-guided Kalman filter predicts object rotation, generates multiple candidate poses, and then obtains the final pose by hierarchical refinement. The 6D pose tracking results guide subsequent 2D tracking and Kalman filter updates, forming a closed-loop system that ensures accurate pose initialization and precise pose tracking. Simulation and real-world experiments demonstrate the effectiveness of our method, achieving real-time and robust 6D pose tracking for fast-moving cameras and objects.", "date": "2025-08-16", "categories": []}
{"id": "2411.06869", "title": "CapeLLM: Support-Free Category-Agnostic Pose Estimation with Multimodal Large Language Models", "abstract": "Category-agnostic pose estimation (CAPE) has traditionally relied on support images with annotated keypoints, a process that is often cumbersome and may fail to fully capture the necessary correspondences across diverse object categories. Recent efforts have explored the use of text queries, leveraging their enhanced stability and generalization capabilities. However, existing approaches often remain constrained by their reliance on support queries, their failure to fully utilize the rich priors embedded in pre-trained large language models, and the limitations imposed by their parametric distribution assumptions. To address these challenges, we introduce CapeLLM, the first multimodal large language model (MLLM) designed for CAPE. Our method only employs query image and detailed text descriptions as an input to estimate category-agnostic keypoints. Our method encompasses effective training strategies and carefully designed instructions for applying the MLLM to CAPE. Moreover, we propose an inference mechanism that further enhances the reasoning process for unseen keypoints. while flexibly modeling their underlying spatial distribution and uncertainty, allowing for adaptive refinement based on contextual cues. We conducted extensive experiments to apply the MLLM to CAPE effectively, focusing not only on the model architecture and prompt design but also on ensuring robustness across input variations. Our approach sets a new state-of-the-art on the MP-100 benchmark in the 1-shot and even 5-shot setting, marking a significant advancement in the field of category-agnostic pose estimation. Code is available at https://github.com/Junhojuno/CapeLLM.", "date": "2024-11-11", "categories": []}
{"id": "2506.01691", "title": "SteerPose: Simultaneous Extrinsic Camera Calibration and Matching from Articulation", "abstract": "Can freely moving humans or animals themselves serve as calibration targets for multi-camera systems while simultaneously estimating their correspondences across views? We humans can solve this problem by mentally rotating the observed 2D poses and aligning them with those in the target views. Inspired by this cognitive ability, we propose SteerPose, a neural network that performs this rotation of 2D poses into another view. By integrating differentiable matching, SteerPose simultaneously performs extrinsic camera calibration and correspondence search within a single unified framework. We also introduce a novel geometric consistency loss that explicitly ensures that the estimated rotation and correspondences result in a valid translation estimation. Experimental results on diverse in-the-wild datasets of humans and animals validate the effectiveness and robustness of the proposed method. Furthermore, we demonstrate that our method can reconstruct the 3D poses of novel animals in multi-camera setups by leveraging off-the-shelf 2D pose estimators and our class-agnostic model.", "date": "2025-06-02", "categories": []}
{"id": "2508.03132", "title": "COFFEE: A Shadow-Resilient Real-Time Pose Estimator for Unknown Tumbling Asteroids using Sparse Neural Networks", "abstract": "The accurate state estimation of unknown bodies in space is a critical challenge with applications ranging from the tracking of space debris to the shape estimation of small bodies. A necessary enabler to this capability is to find and track features on a continuous stream of images. Existing methods, such as SIFT, ORB and AKAZE, achieve real-time but inaccurate pose estimates, whereas modern deep learning methods yield higher quality features at the cost of more demanding computational resources which might not be available on space-qualified hardware. Additionally, both classical and data-driven methods are not robust to the highly opaque self-cast shadows on the object of interest. We show that, as the target body rotates, these shadows may lead to large biases in the resulting pose estimates. For these objects, a bias in the real-time pose estimation algorithm may mislead the spacecraft's state estimator and cause a mission failure, especially if the body undergoes a chaotic tumbling motion. We present COFFEE, the Celestial Occlusion Fast FEature Extractor, a real-time pose estimation framework for asteroids designed to leverage prior information on the sun phase angle given by sun-tracking sensors commonly available onboard spacecraft. By associating salient contours to their projected shadows, a sparse set of features are detected, invariant to the motion of the shadows. A Sparse Neural Network followed by an attention-based Graph Neural Network feature matching model are then jointly trained to provide a set of correspondences between successive frames. The resulting pose estimation pipeline is found to be bias-free, more accurate than classical pose estimation pipelines and an order of magnitude faster than other state-of-the-art deep learning pipelines on synthetic data as well as on renderings of the tumbling asteroid Apophis.", "date": "2025-08-05", "categories": []}
{"id": "2508.03669", "title": "OmniShape: Zero-Shot Multi-Hypothesis Shape and Pose Estimation in the Real World", "abstract": "We would like to estimate the pose and full shape of an object from a single observation, without assuming known 3D model or category. In this work, we propose OmniShape, the first method of its kind to enable probabilistic pose and shape estimation. OmniShape is based on the key insight that shape completion can be decoupled into two multi-modal distributions: one capturing how measurements project into a normalized object reference frame defined by the dataset and the other modelling a prior over object geometries represented as triplanar neural fields. By training separate conditional diffusion models for these two distributions, we enable sampling multiple hypotheses from the joint pose and shape distribution. OmniShape demonstrates compelling performance on challenging real world datasets. Project website: https://tri-ml.github.io/omnishape", "date": "2025-08-05", "categories": []}
{"id": "2410.06694", "title": "OmniPose6D: Towards Short-Term Object Pose Tracking in Dynamic Scenes from Monocular RGB", "abstract": "To address the challenge of short-term object pose tracking in dynamic environments with monocular RGB input, we introduce a large-scale synthetic dataset OmniPose6D, crafted to mirror the diversity of real-world conditions. We additionally present a benchmarking framework for a comprehensive comparison of pose tracking algorithms. We propose a pipeline featuring an uncertainty-aware keypoint refinement network, employing probabilistic modeling to refine pose estimation. Comparative evaluations demonstrate that our approach achieves performance superior to existing baselines on real datasets, underscoring the effectiveness of our synthetic dataset and refinement technique in enhancing tracking precision in dynamic contexts. Our contributions set a new precedent for the development and assessment of object pose tracking methodologies in complex scenes.", "date": "2024-10-09", "categories": []}
{"id": "2508.02157", "title": "Unified Category-Level Object Detection and Pose Estimation from RGB Images using 3D Prototypes", "abstract": "Recognizing objects in images is a fundamental problem in computer vision. Although detecting objects in 2D images is common, many applications require determining their pose in 3D space. Traditional category-level methods rely on RGB-D inputs, which may not always be available, or employ two-stage approaches that use separate models and representations for detection and pose estimation. For the first time, we introduce a unified model that integrates detection and pose estimation into a single framework for RGB images by leveraging neural mesh models with learned features and multi-model RANSAC. Our approach achieves state-of-the-art results for RGB category-level pose estimation on REAL275, improving on the current state-of-the-art by 22.9% averaged across all scale-agnostic metrics. Finally, we demonstrate that our unified method exhibits greater robustness compared to single-stage baselines. Our code and models are available at https://github.com/Fischer-Tom/unified-detection-and-pose-estimation.", "date": "2025-08-04", "categories": []}
{"id": "2409.16178", "title": "SDFit: 3D Object Pose and Shape by Fitting a Morphable SDF to a Single Image", "abstract": "Recovering 3D object pose and shape from a single image is a challenging and ill-posed problem. This is due to strong (self-)occlusions, depth ambiguities, the vast intra- and inter-class shape variance, and the lack of 3D ground truth for natural images. Existing deep-network methods are trained on synthetic datasets to predict 3D shapes, so they often struggle generalizing to real-world images. Moreover, they lack an explicit feedback loop for refining noisy estimates, and primarily focus on geometry without directly considering pixel alignment. To tackle these limitations, we develop a novel render-and-compare optimization framework, called SDFit. This has three key innovations: First, it uses a learned category-specific and morphable signed-distance-function (mSDF) model, and fits this to an image by iteratively refining both 3D pose and shape. The mSDF robustifies inference by constraining the search on the manifold of valid shapes, while allowing for arbitrary shape topologies. Second, SDFit retrieves an initial 3D shape that likely matches the image, by exploiting foundational models for efficient look-up into 3D shape databases. Third, SDFit initializes pose by establishing rich 2D-3D correspondences between the image and the mSDF through foundational features. We evaluate SDFit on three image datasets, i.e., Pix3D, Pascal3D+, and COMIC. SDFit performs on par with SotA feed-forward networks for unoccluded images and common poses, but is uniquely robust to occlusions and uncommon poses. Moreover, it requires no retraining for unseen images. Thus, SDFit contributes new insights for generalizing in the wild. Code is available at https://anticdimi.github.io/sdfit.", "date": "2024-09-24", "categories": []}
{"id": "2403.14559", "title": "VAPO: Visibility-Aware Keypoint Localization for Efficient 6DoF Object Pose Estimation", "abstract": "Localizing predefined 3D keypoints in a 2D image is an effective way to establish 3D-2D correspondences for instance-level 6DoF object pose estimation. However, unreliable localization results of invisible keypoints degrade the quality of correspondences. In this paper, we address this issue by localizing the important keypoints in terms of visibility. Since keypoint visibility information is currently missing in the dataset collection process, we propose an efficient way to generate binary visibility labels from available object-level annotations, for keypoints of both asymmetric objects and symmetric objects. We further derive real-valued visibility-aware importance from binary labels based on the PageRank algorithm. Taking advantage of the flexibility of our visibility-aware importance, we construct VAPO (Visibility-Aware POse estimator) by integrating the visibility-aware importance with a state-of-the-art pose estimation algorithm, along with additional positional encoding. VAPO can work in both CAD-based and CAD-free settings. Extensive experiments are conducted on popular pose estimation benchmarks including Linemod, Linemod-Occlusion, and YCB-V, demonstrating that VAPO clearly achieves state-of-the-art performances. Project page: https://github.com/RuyiLian/VAPO.", "date": "2024-03-21", "categories": []}
{"id": "2502.01312", "title": "CleanPose: Category-Level Object Pose Estimation via Causal Learning and Knowledge Distillation", "abstract": "Category-level object pose estimation aims to recover the rotation, translation and size of unseen instances within predefined categories. In this task, deep neural network-based methods have demonstrated remarkable performance. However, previous studies show they suffer from spurious correlations raised by \"unclean\" confounders in models, hindering their performance on novel instances with significant variations. To address this issue, we propose CleanPose, a novel approach integrating causal learning and knowledge distillation to enhance category-level pose estimation. To mitigate the negative effect of unobserved confounders, we develop a causal inference module based on front-door adjustment, which promotes unbiased estimation by reducing potential spurious correlations. Additionally, to further improve generalization ability, we devise a residual-based knowledge distillation method that has proven effective in providing comprehensive category information guidance. Extensive experiments across multiple benchmarks (REAL275, CAMERA25 and HouseCat6D) hightlight the superiority of proposed CleanPose over state-of-the-art methods. Code will be available at https://github.com/chrislin0621/CleanPose.", "date": "2025-02-03", "categories": []}
{"id": "2507.11077", "title": "GKNet: Graph-based Keypoints Network for Monocular Pose Estimation of Non-cooperative Spacecraft", "abstract": "Monocular pose estimation of non-cooperative spacecraft is significant for on-orbit service (OOS) tasks, such as satellite maintenance, space debris removal, and station assembly. Considering the high demands on pose estimation accuracy, mainstream monocular pose estimation methods typically consist of keypoint detectors and PnP solver. However, current keypoint detectors remain vulnerable to structural symmetry and partial occlusion of non-cooperative spacecraft. To this end, we propose a graph-based keypoints network for the monocular pose estimation of non-cooperative spacecraft, GKNet, which leverages the geometric constraint of keypoints graph. In order to better validate keypoint detectors, we present a moderate-scale dataset for the spacecraft keypoint detection, named SKD, which consists of 3 spacecraft targets, 90,000 simulated images, and corresponding high-precise keypoint annotations. Extensive experiments and an ablation study have demonstrated the high accuracy and effectiveness of our GKNet, compared to the state-of-the-art spacecraft keypoint detectors. The code for GKNet and the SKD dataset is available at https://github.com/Dongzhou-1996/GKNet.", "date": "2025-07-15", "categories": []}
{"id": "2507.06486", "title": "Mask6D: Masked Pose Priors For 6D Object Pose Estimation", "abstract": "Robust 6D object pose estimation in cluttered or occluded conditions using monocular RGB images remains a challenging task. One reason is that current pose estimation networks struggle to extract discriminative, pose-aware features using 2D feature backbones, especially when the available RGB information is limited due to target occlusion in cluttered scenes. To mitigate this, we propose a novel pose estimation-specific pre-training strategy named Mask6D. Our approach incorporates pose-aware 2D-3D correspondence maps and visible mask maps as additional modal information, which is combined with RGB images for the reconstruction-based model pre-training. Essentially, this 2D-3D correspondence maps a transformed 3D object model to 2D pixels, reflecting the pose information of the target in camera coordinate system. Meanwhile, the integrated visible mask map can effectively guide our model to disregard cluttered background information. In addition, an object-focused pre-training loss function is designed to further facilitate our network to remove the background interference. Finally, we fine-tune our pre-trained pose prior-aware network via conventional pose training strategy to realize the reliable pose prediction. Extensive experiments verify that our method outperforms previous end-to-end pose estimation methods.", "date": "2025-07-09", "categories": []}
{"id": "2507.06662", "title": "MK-Pose: Category-Level Object Pose Estimation via Multimodal-Based Keypoint Learning", "abstract": "Category-level object pose estimation, which predicts the pose of objects within a known category without prior knowledge of individual instances, is essential in applications like warehouse automation and manufacturing. Existing methods relying on RGB images or point cloud data often struggle with object occlusion and generalization across different instances and categories. This paper proposes a multimodal-based keypoint learning framework (MK-Pose) that integrates RGB images, point clouds, and category-level textual descriptions. The model uses a self-supervised keypoint detection module enhanced with attention-based query generation, soft heatmap matching and graph-based relational modeling. Additionally, a graph-enhanced feature fusion module is designed to integrate local geometric information and global context. MK-Pose is evaluated on CAMERA25 and REAL275 dataset, and is further tested for cross-dataset capability on HouseCat6D dataset. The results demonstrate that MK-Pose outperforms existing state-of-the-art methods in both IoU and average precision without shape priors. Codes will be released at \\href{https://github.com/yangyifanYYF/MK-Pose}{https://github.com/yangyifanYYF/MK-Pose}.", "date": "2025-07-09", "categories": []}
{"id": "2503.19625", "title": "DynOPETs: A Versatile Benchmark for Dynamic Object Pose Estimation and Tracking in Moving Camera Scenarios", "abstract": "In the realm of object pose estimation, scenarios involving both dynamic objects and moving cameras are prevalent. However, the scarcity of corresponding real-world datasets significantly hinders the development and evaluation of robust pose estimation models. This is largely attributed to the inherent challenges in accurately annotating object poses in dynamic scenes captured by moving cameras. To bridge this gap, this paper presents a novel dataset DynOPETs and a dedicated data acquisition and annotation pipeline tailored for object pose estimation and tracking in such unconstrained environments. Our efficient annotation method innovatively integrates pose estimation and pose tracking techniques to generate pseudo-labels, which are subsequently refined through pose graph optimization. The resulting dataset offers accurate pose annotations for dynamic objects observed from moving cameras. To validate the effectiveness and value of our dataset, we perform comprehensive evaluations using 18 state-of-the-art methods, demonstrating its potential to accelerate research in this challenging domain. The dataset will be made publicly available to facilitate further exploration and advancement in the field.", "date": "2025-03-25", "categories": []}
{"id": "2507.03292", "title": "Zero-shot Inexact CAD Model Alignment from a Single Image", "abstract": "One practical approach to infer 3D scene structure from a single image is to retrieve a closely matching 3D model from a database and align it with the object in the image. Existing methods rely on supervised training with images and pose annotations, which limits them to a narrow set of object categories. To address this, we propose a weakly supervised 9-DoF alignment method for inexact 3D models that requires no pose annotations and generalizes to unseen categories. Our approach derives a novel feature space based on foundation features that ensure multi-view consistency and overcome symmetry ambiguities inherent in foundation features using a self-supervised triplet loss. Additionally, we introduce a texture-invariant pose refinement technique that performs dense alignment in normalized object coordinates, estimated through the enhanced feature space. We conduct extensive evaluations on the real-world ScanNet25k dataset, where our method outperforms SOTA weakly supervised baselines by +4.3% mean alignment accuracy and is the only weakly supervised approach to surpass the supervised ROCA by +2.7%. To assess generalization, we introduce SUN2CAD, a real-world test set with 20 novel object categories, where our method achieves SOTA results without prior training on them.", "date": "2025-07-04", "categories": []}
{"id": "2311.10918", "title": "Jenga Stacking Based on 6D Pose Estimation for Architectural Form Finding Process", "abstract": "This paper includes a review of current state of the art 6d pose estimation methods, as well as a discussion of which pose estimation method should be used in two types of architectural design scenarios. Taking the latest pose estimation research Gen6d as an example, we make a qualitative assessment of the current openset methods in terms of application level, prediction speed, resistance to occlusion, accuracy, resistance to environmental interference, etc. In addition, we try to combine 6D pose estimation and building wind environment assessment to create tangible architectural design approach, we discuss the limitations of the method and point out the direction in which 6d pose estimation is eager to progress in this scenario.", "date": "2023-11-18", "categories": []}
{"id": "2507.01478", "title": "Active Control Points-based 6DoF Pose Tracking for Industrial Metal Objects", "abstract": "Visual pose tracking is playing an increasingly vital role in industrial contexts in recent years. However, the pose tracking for industrial metal objects remains a challenging task especially in the real world-environments, due to the reflection characteristic of metal objects. To address this issue, we propose a novel 6DoF pose tracking method based on active control points. The method uses image control points to generate edge feature for optimization actively instead of 6DoF pose-based rendering, and serve them as optimization variables. We also introduce an optimal control point regression method to improve robustness. The proposed tracking method performs effectively in both dataset evaluation and real world tasks, providing a viable solution for real-time tracking of industrial metal objects. Our source code is made publicly available at: https://github.com/tomatoma00/ACPTracking.", "date": "2025-07-02", "categories": []}
{"id": "2507.00224", "title": "Computer Vision for Objects used in Group Work: Challenges and Opportunities", "abstract": "Interactive and spatially aware technologies are transforming educational frameworks, particularly in K-12 settings where hands-on exploration fosters deeper conceptual understanding. However, during collaborative tasks, existing systems often lack the ability to accurately capture real-world interactions between students and physical objects. This issue could be addressed with automatic 6D pose estimation, i.e., estimation of an object's position and orientation in 3D space from RGB images or videos. For collaborative groups that interact with physical objects, 6D pose estimates allow AI systems to relate objects and entities. As part of this work, we introduce FiboSB, a novel and challenging 6D pose video dataset featuring groups of three participants solving an interactive task featuring small hand-held cubes and a weight scale. This setup poses unique challenges for 6D pose because groups are holistically recorded from a distance in order to capture all participants -- this, coupled with the small size of the cubes, makes 6D pose estimation inherently non-trivial. We evaluated four state-of-the-art 6D pose estimation methods on FiboSB, exposing the limitations of current algorithms on collaborative group work. An error analysis of these methods reveals that the 6D pose methods' object detection modules fail. We address this by fine-tuning YOLO11-x for FiboSB, achieving an overall mAP_50 of 0.898. The dataset, benchmark results, and analysis of YOLO11-x errors presented here lay the groundwork for leveraging the estimation of 6D poses in difficult collaborative contexts.", "date": "2025-06-30", "categories": []}
{"id": "2507.00984", "title": "Box Pose and Shape Estimation and Domain Adaptation for Large-Scale Warehouse Automation", "abstract": "Modern warehouse automation systems rely on fleets of intelligent robots that generate vast amounts of data -- most of which remains unannotated. This paper develops a self-supervised domain adaptation pipeline that leverages real-world, unlabeled data to improve perception models without requiring manual annotations. Our work focuses specifically on estimating the pose and shape of boxes and presents a correct-and-certify pipeline for self-supervised box pose and shape estimation. We extensively evaluate our approach across a range of simulated and real industrial settings, including adaptation to a large-scale real-world dataset of 50,000 images. The self-supervised model significantly outperforms models trained solely in simulation and shows substantial improvements over a zero-shot 3D bounding box estimation baseline.", "date": "2025-07-01", "categories": []}
{"id": "2506.22720", "title": "Deterministic Object Pose Confidence Region Estimation", "abstract": "6D pose confidence region estimation has emerged as a critical direction, aiming to perform uncertainty quantification for assessing the reliability of estimated poses. However, current sampling-based approach suffers from critical limitations that severely impede their practical deployment: 1) the sampling speed significantly decreases as the number of samples increases. 2) the derived confidence regions are often excessively large. To address these challenges, we propose a deterministic and efficient method for estimating pose confidence regions. Our approach uses inductive conformal prediction to calibrate the deterministically regressed Gaussian keypoint distributions into 2D keypoint confidence regions. We then leverage the implicit function theorem to propagate these keypoint confidence regions directly into 6D pose confidence regions. This method avoids the inefficiency and inflated region sizes associated with sampling and ensembling. It provides compact confidence regions that cover the ground-truth poses with a user-defined confidence level. Experimental results on the LineMOD Occlusion and SPEED datasets show that our method achieves higher pose estimation accuracy with reduced computational time. For the same coverage rate, our method yields significantly smaller confidence region volumes, reducing them by up to 99.9\\% for rotations and 99.8\\% for translations. The code will be available soon.", "date": "2025-06-28", "categories": []}
{"id": "2502.04293", "title": "GCE-Pose: Global Context Enhancement for Category-level Object Pose Estimation", "abstract": "A key challenge in model-free category-level pose estimation is the extraction of contextual object features that generalize across varying instances within a specific category. Recent approaches leverage foundational features to capture semantic and geometry cues from data. However, these approaches fail under partial visibility. We overcome this with a first-complete-then-aggregate strategy for feature extraction utilizing class priors. In this paper, we present GCE-Pose, a method that enhances pose estimation for novel instances by integrating category-level global context prior. GCE-Pose performs semantic shape reconstruction with a proposed Semantic Shape Reconstruction (SSR) module. Given an unseen partial RGB-D object instance, our SSR module reconstructs the instance's global geometry and semantics by deforming category-specific 3D semantic prototypes through a learned deep Linear Shape Model. We further introduce a Global Context Enhanced (GCE) feature fusion module that effectively fuses features from partial RGB-D observations and the reconstructed global context. Extensive experiments validate the impact of our global context prior and the effectiveness of the GCE fusion module, demonstrating that GCE-Pose significantly outperforms existing methods on challenging real-world datasets HouseCat6D and NOCS-REAL275. Our project page is available at https://colin-de.github.io/GCE-Pose/.", "date": "2025-02-06", "categories": []}
{"id": "2506.18856", "title": "RAG-6DPose: Retrieval-Augmented 6D Pose Estimation via Leveraging CAD as Knowledge Base", "abstract": "Accurate 6D pose estimation is key for robotic manipulation, enabling precise object localization for tasks like grasping. We present RAG-6DPose, a retrieval-augmented approach that leverages 3D CAD models as a knowledge base by integrating both visual and geometric cues. Our RAG-6DPose roughly contains three stages: 1) Building a Multi-Modal CAD Knowledge Base by extracting 2D visual features from multi-view CAD rendered images and also attaching 3D points; 2) Retrieving relevant CAD features from the knowledge base based on the current query image via our ReSPC module; and 3) Incorporating retrieved CAD information to refine pose predictions via retrieval-augmented decoding. Experimental results on standard benchmarks and real-world robotic tasks demonstrate the effectiveness and robustness of our approach, particularly in handling occlusions and novel viewpoints. Supplementary material is available on our project website: https://sressers.github.io/RAG-6DPose .", "date": "2025-06-23", "categories": []}
{"id": "2506.17119", "title": "RGBTrack: Fast, Robust Depth-Free 6D Pose Estimation and Tracking", "abstract": "We introduce a robust framework, RGBTrack, for real-time 6D pose estimation and tracking that operates solely on RGB data, thereby eliminating the need for depth input for such dynamic and precise object pose tracking tasks. Building on the FoundationPose architecture, we devise a novel binary search strategy combined with a render-and-compare mechanism to efficiently infer depth and generate robust pose hypotheses from true-scale CAD models. To maintain stable tracking in dynamic scenarios, including rapid movements and occlusions, RGBTrack integrates state-of-the-art 2D object tracking (XMem) with a Kalman filter and a state machine for proactive object pose recovery. In addition, RGBTrack's scale recovery module dynamically adapts CAD models of unknown scale using an initial depth estimate, enabling seamless integration with modern generative reconstruction techniques. Extensive evaluations on benchmark datasets demonstrate that RGBTrack's novel depth-free approach achieves competitive accuracy and real-time performance, making it a promising practical solution candidate for application areas including robotics, augmented reality, and computer vision.\n  The source code for our implementation will be made publicly available at https://github.com/GreatenAnoymous/RGBTrack.git.", "date": "2025-06-20", "categories": []}
{"id": "2408.08234", "title": "Comparative Evaluation of 3D Reconstruction Methods for Object Pose Estimation", "abstract": "Object pose estimation is essential to many industrial applications involving robotic manipulation, navigation, and augmented reality. Current generalizable object pose estimators, i.e., approaches that do not need to be trained per object, rely on accurate 3D models. Predominantly, CAD models are used, which can be hard to obtain in practice. At the same time, it is often possible to acquire images of an object. Naturally, this leads to the question whether 3D models reconstructed from images are sufficient to facilitate accurate object pose estimation. We aim to answer this question by proposing a novel benchmark for measuring the impact of 3D reconstruction quality on pose estimation accuracy. Our benchmark provides calibrated images for object reconstruction registered with the test images of the YCB-V dataset for pose evaluation under the BOP benchmark format. Detailed experiments with multiple state-of-the-art 3D reconstruction and object pose estimation approaches show that the geometry produced by modern reconstruction methods is often sufficient for accurate pose estimation. Our experiments lead to interesting observations: (1) Standard metrics for measuring 3D reconstruction quality are not necessarily indicative of pose estimation accuracy, which shows the need for dedicated benchmarks such as ours. (2) Classical, non-learning-based approaches can perform on par with modern learning-based reconstruction techniques and can even offer a better reconstruction time-pose accuracy tradeoff. (3) There is still a sizable gap between performance with reconstructed and with CAD models. To foster research on closing this gap, our benchmark is publicly available at https://github.com/VarunBurde/reconstruction_pose_benchmark}.", "date": "2024-08-15", "categories": []}
{"id": "2506.00599", "title": "XYZ-IBD: A High-precision Bin-picking Dataset for Object 6D Pose Estimation Capturing Real-world Industrial Complexity", "abstract": "We introduce XYZ-IBD, a bin-picking dataset for 6D pose estimation that captures real-world industrial complexity, including challenging object geometries, reflective materials, severe occlusions, and dense clutter. The dataset reflects authentic robotic manipulation scenarios with millimeter-accurate annotations. Unlike existing datasets that primarily focus on household objects, which approach saturation,XYZ-IBD represents the unsolved realistic industrial conditions. The dataset features 15 texture-less, metallic, and mostly symmetrical objects of varying shapes and sizes. These objects are heavily occluded and randomly arranged in bins with high density, replicating the challenges of real-world bin-picking. XYZ-IBD was collected using two high-precision industrial cameras and one commercially available camera, providing RGB, grayscale, and depth images. It contains 75 multi-view real-world scenes, along with a large-scale synthetic dataset rendered under simulated bin-picking conditions. We employ a meticulous annotation pipeline that includes anti-reflection spray, multi-view depth fusion, and semi-automatic annotation, achieving millimeter-level pose labeling accuracy required for industrial manipulation. Quantification in simulated environments confirms the reliability of the ground-truth annotations. We benchmark state-of-the-art methods on 2D detection, 6D pose estimation, and depth estimation tasks on our dataset, revealing significant performance degradation in our setups compared to current academic household benchmarks. By capturing the complexity of real-world bin-picking scenarios, XYZ-IBD introduces more realistic and challenging problems for future research. The dataset and benchmark are publicly available at https://xyz-ibd.github.io/XYZ-IBD/.", "date": "2025-05-31", "categories": []}
{"id": "2506.12782", "title": "A large-scale, physically-based synthetic dataset for satellite pose estimation", "abstract": "The Deep Learning Visual Space Simulation System (DLVS3) introduces a novel synthetic dataset generator and a simulation pipeline specifically designed for training and testing satellite pose estimation solutions. This work introduces the DLVS3-HST-V1 dataset, which focuses on the Hubble Space Telescope (HST) as a complex, articulated target. The dataset is generated using advanced real-time and offline rendering technologies, integrating high-fidelity 3D models, dynamic lighting (including secondary sources like Earth reflection), and physically accurate material properties. The pipeline supports the creation of large-scale, richly annotated image sets with ground-truth 6-DoF pose and keypoint data, semantic segmentation, depth, and normal maps. This enables the training and benchmarking of deep learning-based pose estimation solutions under realistic, diverse, and challenging visual conditions. The paper details the dataset generation process, the simulation architecture, and the integration with deep learning frameworks, and positions DLVS3 as a significant step toward closing the domain gap for autonomous spacecraft operations in proximity and servicing missions.", "date": "2025-06-15", "categories": []}
{"id": "2506.09699", "title": "CHIP: A multi-sensor dataset for 6D pose estimation of chairs in industrial settings", "abstract": "Accurate 6D pose estimation of complex objects in 3D environments is essential for effective robotic manipulation. Yet, existing benchmarks fall short in evaluating 6D pose estimation methods under realistic industrial conditions, as most datasets focus on household objects in domestic settings, while the few available industrial datasets are limited to artificial setups with objects placed on tables. To bridge this gap, we introduce CHIP, the first dataset designed for 6D pose estimation of chairs manipulated by a robotic arm in a real-world industrial environment. CHIP includes seven distinct chairs captured using three different RGBD sensing technologies and presents unique challenges, such as distractor objects with fine-grained differences and severe occlusions caused by the robotic arm and human operators. CHIP comprises 77,811 RGBD images annotated with ground-truth 6D poses automatically derived from the robot's kinematics, averaging 11,115 annotations per chair. We benchmark CHIP using three zero-shot 6D pose estimation methods, assessing performance across different sensor types, localization priors, and occlusion levels. Results show substantial room for improvement, highlighting the unique challenges posed by the dataset. CHIP will be publicly released.", "date": "2025-06-11", "categories": []}
{"id": "2506.09784", "title": "Accurate and efficient zero-shot 6D pose estimation with frozen foundation models", "abstract": "Estimating the 6D pose of objects from RGBD data is a fundamental problem in computer vision, with applications in robotics and augmented reality. A key challenge is achieving generalization to novel objects that were not seen during training. Most existing approaches address this by scaling up training on synthetic data tailored to the task, a process that demands substantial computational resources. But is task-specific training really necessary for accurate and efficient 6D pose estimation of novel objects? To answer No!, we introduce FreeZeV2, the second generation of FreeZe: a training-free method that achieves strong generalization to unseen objects by leveraging geometric and vision foundation models pre-trained on unrelated data. FreeZeV2 improves both accuracy and efficiency over FreeZe through three key contributions: (i) a sparse feature extraction strategy that reduces inference-time computation without sacrificing accuracy; (ii) a feature-aware scoring mechanism that improves both pose selection during RANSAC-based 3D registration and the final ranking of pose candidates; and (iii) a modular design that supports ensembles of instance segmentation models, increasing robustness to segmentation masks errors. We evaluate FreeZeV2 on the seven core datasets of the BOP Benchmark, where it establishes a new state-of-the-art in 6D pose estimation of unseen objects. When using the same segmentation masks, FreeZeV2 achieves a remarkable 8x speedup over FreeZe while also improving accuracy by 5%. When using ensembles of segmentation models, FreeZeV2 gains an additional 8% in accuracy while still running 2.5x faster than FreeZe. FreeZeV2 was awarded Best Overall Method at the BOP Challenge 2024.", "date": "2025-06-11", "categories": []}
{"id": "2503.20220", "title": "DINeMo: Learning Neural Mesh Models with no 3D Annotations", "abstract": "Category-level 3D/6D pose estimation is a crucial step towards comprehensive 3D scene understanding, which would enable a broad range of applications in robotics and embodied AI. Recent works explored neural mesh models that approach a range of 2D and 3D tasks from an analysis-by-synthesis perspective. Despite the largely enhanced robustness to partial occlusion and domain shifts, these methods depended heavily on 3D annotations for part-contrastive learning, which confines them to a narrow set of categories and hinders efficient scaling. In this work, we present DINeMo, a novel neural mesh model that is trained with no 3D annotations by leveraging pseudo-correspondence obtained from large visual foundation models. We adopt a bidirectional pseudo-correspondence generation method, which produce pseudo correspondence utilize both local appearance features and global context information. Experimental results on car datasets demonstrate that our DINeMo outperforms previous zero- and few-shot 3D pose estimation by a wide margin, narrowing the gap with fully-supervised methods by 67.3%. Our DINeMo also scales effectively and efficiently when incorporating more unlabeled images during training, which demonstrate the advantages over supervised learning methods that rely on 3D annotations. Our project page is available at https://analysis-by-synthesis.github.io/DINeMo/.", "date": "2025-03-26", "categories": []}
{"id": "2506.07155", "title": "GoTrack: Generic 6DoF Object Pose Refinement and Tracking", "abstract": "We introduce GoTrack, an efficient and accurate CAD-based method for 6DoF object pose refinement and tracking, which can handle diverse objects without any object-specific training. Unlike existing tracking methods that rely solely on an analysis-by-synthesis approach for model-to-frame registration, GoTrack additionally integrates frame-to-frame registration, which saves compute and stabilizes tracking. Both types of registration are realized by optical flow estimation. The model-to-frame registration is noticeably simpler than in existing methods, relying only on standard neural network blocks (a transformer is trained on top of DINOv2) and producing reliable pose confidence scores without a scoring network. For the frame-to-frame registration, which is an easier problem as consecutive video frames are typically nearly identical, we employ a light off-the-shelf optical flow model. We demonstrate that GoTrack can be seamlessly combined with existing coarse pose estimation methods to create a minimal pipeline that reaches state-of-the-art RGB-only results on standard benchmarks for 6DoF object pose estimation and tracking. Our source code and trained models are publicly available at https://github.com/facebookresearch/gotrack", "date": "2025-06-08", "categories": []}
{"id": "2506.07996", "title": "UA-Pose: Uncertainty-Aware 6D Object Pose Estimation and Online Object Completion with Partial References", "abstract": "6D object pose estimation has shown strong generalizability to novel objects. However, existing methods often require either a complete, well-reconstructed 3D model or numerous reference images that fully cover the object. Estimating 6D poses from partial references, which capture only fragments of an object's appearance and geometry, remains challenging. To address this, we propose UA-Pose, an uncertainty-aware approach for 6D object pose estimation and online object completion specifically designed for partial references. We assume access to either (1) a limited set of RGBD images with known poses or (2) a single 2D image. For the first case, we initialize a partial object 3D model based on the provided images and poses, while for the second, we use image-to-3D techniques to generate an initial object 3D model. Our method integrates uncertainty into the incomplete 3D model, distinguishing between seen and unseen regions. This uncertainty enables confidence assessment in pose estimation and guides an uncertainty-aware sampling strategy for online object completion, enhancing robustness in pose estimation accuracy and improving object completeness. We evaluate our method on the YCB-Video, YCBInEOAT, and HO3D datasets, including RGBD sequences of YCB objects manipulated by robots and human hands. Experimental results demonstrate significant performance improvements over existing methods, particularly when object observations are incomplete or partially captured. Project page: https://minfenli.github.io/UA-Pose/", "date": "2025-06-09", "categories": []}
{"id": "2506.05719", "title": "You Only Estimate Once: Unified, One-stage, Real-Time Category-level Articulated Object 6D Pose Estimation for Robotic Grasping", "abstract": "This paper addresses the problem of category-level pose estimation for articulated objects in robotic manipulation tasks. Recent works have shown promising results in estimating part pose and size at the category level. However, these approaches primarily follow a complex multi-stage pipeline that first segments part instances in the point cloud and then estimates the Normalized Part Coordinate Space (NPCS) representation for 6D poses. These approaches suffer from high computational costs and low performance in real-time robotic tasks. To address these limitations, we propose YOEO, a single-stage method that simultaneously outputs instance segmentation and NPCS representations in an end-to-end manner. We use a unified network to generate point-wise semantic labels and centroid offsets, allowing points from the same part instance to vote for the same centroid. We further utilize a clustering algorithm to distinguish points based on their estimated centroid distances. Finally, we first separate the NPCS region of each instance. Then, we align the separated regions with the real point cloud to recover the final pose and size. Experimental results on the GAPart dataset demonstrate the pose estimation capabilities of our proposed single-shot method. We also deploy our synthetically-trained model in a real-world setting, providing real-time visual feedback at 200Hz, enabling a physical Kinova robot to interact with unseen articulated objects. This showcases the utility and effectiveness of our proposed method.", "date": "2025-06-06", "categories": []}
{"id": "2309.13570", "title": "Robust 6DoF Pose Estimation Against Depth Noise and a Comprehensive Evaluation on a Mobile Dataset", "abstract": "Robust 6DoF pose estimation with mobile devices is the foundation for applications in robotics, augmented reality, and digital twin localization. In this paper, we extensively investigate the robustness of existing RGBD-based 6DoF pose estimation methods against varying levels of depth sensor noise. We highlight that existing 6DoF pose estimation methods suffer significant performance discrepancies due to depth measurement inaccuracies. In response to the robustness issue, we present a simple and effective transformer-based 6DoF pose estimation approach called DTTDNet, featuring a novel geometric feature filtering module and a Chamfer distance loss for training. Moreover, we advance the field of robust 6DoF pose estimation and introduce a new dataset -- Digital Twin Tracking Dataset Mobile (DTTD-Mobile), tailored for digital twin object tracking with noisy depth data from the mobile RGBD sensor suite of the Apple iPhone 14 Pro. Extensive experiments demonstrate that DTTDNet significantly outperforms state-of-the-art methods at least 4.32, up to 60.74 points in ADD metrics on the DTTD-Mobile. More importantly, our approach exhibits superior robustness to varying levels of measurement noise, setting a new benchmark for robustness to measurement noise. The project page is publicly available at https://openark-berkeley.github.io/DTTDNet/.", "date": "2023-09-24", "categories": []}
{"id": "2504.11230", "title": "CAP-Net: A Unified Network for 6D Pose and Size Estimation of Categorical Articulated Parts from a Single RGB-D Image", "abstract": "This paper tackles category-level pose estimation of articulated objects in robotic manipulation tasks and introduces a new benchmark dataset. While recent methods estimate part poses and sizes at the category level, they often rely on geometric cues and complex multi-stage pipelines that first segment parts from the point cloud, followed by Normalized Part Coordinate Space (NPCS) estimation for 6D poses. These approaches overlook dense semantic cues from RGB images, leading to suboptimal accuracy, particularly for objects with small parts. To address these limitations, we propose a single-stage Network, CAP-Net, for estimating the 6D poses and sizes of Categorical Articulated Parts. This method combines RGB-D features to generate instance segmentation and NPCS representations for each part in an end-to-end manner. CAP-Net uses a unified network to simultaneously predict point-wise class labels, centroid offsets, and NPCS maps. A clustering algorithm then groups points of the same predicted class based on their estimated centroid distances to isolate each part. Finally, the NPCS region of each part is aligned with the point cloud to recover its final pose and size. To bridge the sim-to-real domain gap, we introduce the RGBD-Art dataset, the largest RGB-D articulated dataset to date, featuring photorealistic RGB images and depth noise simulated from real sensors. Experimental evaluations on the RGBD-Art dataset demonstrate that our method significantly outperforms the state-of-the-art approach. Real-world deployments of our model in robotic tasks underscore its robustness and exceptional sim-to-real transfer capabilities, confirming its substantial practical utility. Our dataset, code and pre-trained models are available on the project page.", "date": "2025-04-15", "categories": []}
{"id": "2505.24636", "title": "Category-Level 6D Object Pose Estimation in Agricultural Settings Using a Lattice-Deformation Framework and Diffusion-Augmented Synthetic Data", "abstract": "Accurate 6D object pose estimation is essential for robotic grasping and manipulation, particularly in agriculture, where fruits and vegetables exhibit high intra-class variability in shape, size, and texture. The vast majority of existing methods rely on instance-specific CAD models or require depth sensors to resolve geometric ambiguities, making them impractical for real-world agricultural applications. In this work, we introduce PLANTPose, a novel framework for category-level 6D pose estimation that operates purely on RGB input. PLANTPose predicts both the 6D pose and deformation parameters relative to a base mesh, allowing a single category-level CAD model to adapt to unseen instances. This enables accurate pose estimation across varying shapes without relying on instance-specific data. To enhance realism and improve generalization, we also leverage Stable Diffusion to refine synthetic training images with realistic texturing, mimicking variations due to ripeness and environmental factors and bridging the domain gap between synthetic data and the real world. Our evaluations on a challenging benchmark that includes bananas of various shapes, sizes, and ripeness status demonstrate the effectiveness of our framework in handling large intraclass variations while maintaining accurate 6D pose predictions, significantly outperforming the state-of-the-art RGB-based approach MegaPose.", "date": "2025-05-30", "categories": []}
{"id": "2505.24669", "title": "6D Pose Estimation on Point Cloud Data through Prior Knowledge Integration: A Case Study in Autonomous Disassembly", "abstract": "The accurate estimation of 6D pose remains a challenging task within the computer vision domain, even when utilizing 3D point cloud data. Conversely, in the manufacturing domain, instances arise where leveraging prior knowledge can yield advancements in this endeavor. This study focuses on the disassembly of starter motors to augment the engineering of product life cycles. A pivotal objective in this context involves the identification and 6D pose estimation of bolts affixed to the motors, facilitating automated disassembly within the manufacturing workflow. Complicating matters, the presence of occlusions and the limitations of single-view data acquisition, notably when motors are placed in a clamping system, obscure certain portions and render some bolts imperceptible. Consequently, the development of a comprehensive pipeline capable of acquiring complete bolt information is imperative to avoid oversight in bolt detection. In this paper, employing the task of bolt detection within the scope of our project as a pertinent use case, we introduce a meticulously devised pipeline. This multi-stage pipeline effectively captures the 6D information with regard to all bolts on the motor, thereby showcasing the effective utilization of prior knowledge in handling this challenging task. The proposed methodology not only contributes to the field of 6D pose estimation but also underscores the viability of integrating domain-specific insights to tackle complex problems in manufacturing and automation.", "date": "2025-05-30", "categories": []}
{"id": "2505.10841", "title": "RefPose: Leveraging Reference Geometric Correspondences for Accurate 6D Pose Estimation of Unseen Objects", "abstract": "Estimating the 6D pose of unseen objects from monocular RGB images remains a challenging problem, especially due to the lack of prior object-specific knowledge. To tackle this issue, we propose RefPose, an innovative approach to object pose estimation that leverages a reference image and geometric correspondence as guidance. RefPose first predicts an initial pose by using object templates to render the reference image and establish the geometric correspondence needed for the refinement stage. During the refinement stage, RefPose estimates the geometric correspondence of the query based on the generated references and iteratively refines the pose through a render-and-compare approach. To enhance this estimation, we introduce a correlation volume-guided attention mechanism that effectively captures correlations between the query and reference images. Unlike traditional methods that depend on pre-defined object models, RefPose dynamically adapts to new object shapes by leveraging a reference image and geometric correspondence. This results in robust performance across previously unseen objects. Extensive evaluation on the BOP benchmark datasets shows that RefPose achieves state-of-the-art results while maintaining a competitive runtime.", "date": "2025-05-16", "categories": []}
{"id": "2505.11439", "title": "SurgPose: Generalisable Surgical Instrument Pose Estimation using Zero-Shot Learning and Stereo Vision", "abstract": "Accurate pose estimation of surgical tools in Robot-assisted Minimally Invasive Surgery (RMIS) is essential for surgical navigation and robot control. While traditional marker-based methods offer accuracy, they face challenges with occlusions, reflections, and tool-specific designs. Similarly, supervised learning methods require extensive training on annotated datasets, limiting their adaptability to new tools. Despite their success in other domains, zero-shot pose estimation models remain unexplored in RMIS for pose estimation of surgical instruments, creating a gap in generalising to unseen surgical tools. This paper presents a novel 6 Degrees of Freedom (DoF) pose estimation pipeline for surgical instruments, leveraging state-of-the-art zero-shot RGB-D models like the FoundationPose and SAM-6D. We advanced these models by incorporating vision-based depth estimation using the RAFT-Stereo method, for robust depth estimation in reflective and textureless environments. Additionally, we enhanced SAM-6D by replacing its instance segmentation module, Segment Anything Model (SAM), with a fine-tuned Mask R-CNN, significantly boosting segmentation accuracy in occluded and complex conditions. Extensive validation reveals that our enhanced SAM-6D surpasses FoundationPose in zero-shot pose estimation of unseen surgical instruments, setting a new benchmark for zero-shot RGB-D pose estimation in RMIS. This work enhances the generalisability of pose estimation for unseen objects and pioneers the application of RGB-D zero-shot methods in RMIS.", "date": "2025-05-16", "categories": []}
{"id": "2505.04109", "title": "One2Any: One-Reference 6D Pose Estimation for Any Object", "abstract": "  6D object pose estimation remains challenging for many applications due to\ndependencies on complete 3D models, multi-view images, or training limited to\nspecific object categories. These requirements make generalization to novel\nobjects difficult for which neither 3D models nor multi-view images may be\navailable. To address this, we propose a novel method One2Any that estimates\nthe relative 6-degrees of freedom (DOF) object pose using only a single\nreference-single query RGB-D image, without prior knowledge of its 3D model,\nmulti-view data, or category constraints. We treat object pose estimation as an\nencoding-decoding process, first, we obtain a comprehensive Reference Object\nPose Embedding (ROPE) that encodes an object shape, orientation, and texture\nfrom a single reference view. Using this embedding, a U-Net-based pose decoding\nmodule produces Reference Object Coordinate (ROC) for new views, enabling fast\nand accurate pose estimation. This simple encoding-decoding framework allows\nour model to be trained on any pair-wise pose data, enabling large-scale\ntraining and demonstrating great scalability. Experiments on multiple benchmark\ndatasets demonstrate that our model generalizes well to novel objects,\nachieving state-of-the-art accuracy and robustness even rivaling methods that\nrequire multi-view or CAD inputs, at a fraction of compute.\n", "date": "2025-05-07", "categories": []}
{"id": "2505.02335", "title": "6D Pose Estimation on Spoons and Hands", "abstract": "  Accurate dietary monitoring is essential for promoting healthier eating\nhabits. A key area of research is how people interact and consume food using\nutensils and hands. By tracking their position and orientation, it is possible\nto estimate the volume of food being consumed, or monitor eating behaviours,\nhighly useful insights into nutritional intake that can be more reliable than\npopular methods such as self-reporting. Hence, this paper implements a system\nthat analyzes stationary video feed of people eating, using 6D pose estimation\nto track hand and spoon movements to capture spatial position and orientation.\nIn doing so, we examine the performance of two state-of-the-art (SOTA) video\nobject segmentation (VOS) models, both quantitatively and qualitatively, and\nidentify main sources of error within the system.\n", "date": "2025-05-05", "categories": []}
{"id": "2505.02501", "title": "Corr2Distrib: Making Ambiguous Correspondences an Ally to Predict\n  Reliable 6D Pose Distributions", "abstract": "  We introduce Corr2Distrib, the first correspondence-based method which\nestimates a 6D camera pose distribution from an RGB image, explaining the\nobservations. Indeed, symmetries and occlusions introduce visual ambiguities,\nleading to multiple valid poses. While a few recent methods tackle this\nproblem, they do not rely on local correspondences which, according to the BOP\nChallenge, are currently the most effective way to estimate a single 6DoF pose\nsolution. Using correspondences to estimate a pose distribution is not\nstraightforward, since ambiguous correspondences induced by visual ambiguities\ndrastically decrease the performance of PnP. With Corr2Distrib, we turn these\nambiguities into an advantage to recover all valid poses. Corr2Distrib first\nlearns a symmetry-aware representation for each 3D point on the object's\nsurface, characterized by a descriptor and a local frame. This representation\nenables the generation of 3DoF rotation hypotheses from single 2D-3D\ncorrespondences. Next, we refine these hypotheses into a 6DoF pose distribution\nusing PnP and pose scoring. Our experimental evaluations on complex\nnon-synthetic scenes show that Corr2Distrib outperforms state-of-the-art\nsolutions for both pose distribution estimation and single pose estimation from\nan RGB image, demonstrating the potential of correspondences-based approaches.\n", "date": "2025-05-05", "categories": []}
{"id": "2504.19572", "title": "Category-Level and Open-Set Object Pose Estimation for Robotics", "abstract": "  Object pose estimation enables a variety of tasks in computer vision and\nrobotics, including scene understanding and robotic grasping. The complexity of\na pose estimation task depends on the unknown variables related to the target\nobject. While instance-level methods already excel for opaque and Lambertian\nobjects, category-level and open-set methods, where texture, shape, and size\nare partially or entirely unknown, still struggle with these basic material\nproperties. Since texture is unknown in these scenarios, it cannot be used for\ndisambiguating object symmetries, another core challenge of 6D object pose\nestimation. The complexity of estimating 6D poses with such a manifold of\nunknowns led to various datasets, accuracy metrics, and algorithmic solutions.\nThis paper compares datasets, accuracy metrics, and algorithms for solving 6D\npose estimation on the category-level. Based on this comparison, we analyze how\nto bridge category-level and open-set object pose estimation to reach\ngeneralization and provide actionable recommendations.\n", "date": "2025-04-28", "categories": []}
{"id": "2504.02812", "title": "BOP Challenge 2024 on Model-Based and Model-Free 6D Object Pose\n  Estimation", "abstract": "  We present the evaluation methodology, datasets and results of the BOP\nChallenge 2024, the 6th in a series of public competitions organized to capture\nthe state of the art in 6D object pose estimation and related tasks. In 2024,\nour goal was to transition BOP from lab-like setups to real-world scenarios.\nFirst, we introduced new model-free tasks, where no 3D object models are\navailable and methods need to onboard objects just from provided reference\nvideos. Second, we defined a new, more practical 6D object detection task where\nidentities of objects visible in a test image are not provided as input. Third,\nwe introduced new BOP-H3 datasets recorded with high-resolution sensors and\nAR/VR headsets, closely resembling real-world scenarios. BOP-H3 include 3D\nmodels and onboarding videos to support both model-based and model-free tasks.\nParticipants competed on seven challenge tracks. Notably, the best 2024 method\nfor model-based 6D localization of unseen objects (FreeZeV2.1) achieves 22%\nhigher accuracy on BOP-Classic-Core than the best 2023 method (GenFlow), and is\nonly 4% behind the best 2023 method for seen objects (GPose2023) although being\nsignificantly slower (24.9 vs 2.7s per image). A more practical 2024 method for\nthis task is Co-op which takes only 0.8s per image and is 13% more accurate\nthan GenFlow. Methods have similar rankings on 6D detection as on 6D\nlocalization but higher run time. On model-based 2D detection of unseen\nobjects, the best 2024 method (MUSE) achieves 21--29% relative improvement\ncompared to the best 2023 method (CNOS). However, the 2D detection accuracy for\nunseen objects is still -35% behind the accuracy for seen objects (GDet2023),\nand the 2D detection stage is consequently the main bottleneck of existing\npipelines for 6D localization/detection of unseen objects. The online\nevaluation system stays open and is available at http://bop.felk.cvut.cz/\n", "date": "2025-04-03", "categories": []}
{"id": "2504.15329", "title": "Vision6D: 3D-to-2D Interactive Visualization and Annotation Tool for 6D\n  Pose Estimation", "abstract": "  Accurate 6D pose estimation has gained more attention over the years for\nrobotics-assisted tasks that require precise interaction with physical objects.\nThis paper presents an interactive 3D-to-2D visualization and annotation tool\nto support the 6D pose estimation research community. To the best of our\nknowledge, the proposed work is the first tool that allows users to visualize\nand manipulate 3D objects interactively on a 2D real-world scene, along with a\ncomprehensive user study. This system supports robust 6D camera pose annotation\nby providing both visual cues and spatial relationships to determine object\nposition and orientation in various environments. The annotation feature in\nVision6D is particularly helpful in scenarios where the transformation matrix\nbetween the camera and world objects is unknown, as it enables accurate\nannotation of these objects' poses using only the camera intrinsic matrix. This\ncapability serves as a foundational step in developing and training advanced\npose estimation models across various domains. We evaluate Vision6D's\neffectiveness by utilizing widely-used open-source pose estimation datasets\nLinemod and HANDAL through comparisons between the default ground-truth camera\nposes with manual annotations. A user study was performed to show that Vision6D\ngenerates accurate pose annotations via visual cues in an intuitive 3D user\ninterface. This approach aims to bridge the gap between 2D scene projections\nand 3D scenes, offering an effective way for researchers and developers to\nsolve 6D pose annotation related problems. The software is open-source and\npublicly available at https://github.com/InteractiveGL/vision6D.\n", "date": "2025-04-21", "categories": []}
{"id": "2504.13179", "title": "ViTa-Zero: Zero-shot Visuotactile Object 6D Pose Estimation", "abstract": "  Object 6D pose estimation is a critical challenge in robotics, particularly\nfor manipulation tasks. While prior research combining visual and tactile\n(visuotactile) information has shown promise, these approaches often struggle\nwith generalization due to the limited availability of visuotactile data. In\nthis paper, we introduce ViTa-Zero, a zero-shot visuotactile pose estimation\nframework. Our key innovation lies in leveraging a visual model as its backbone\nand performing feasibility checking and test-time optimization based on\nphysical constraints derived from tactile and proprioceptive observations.\nSpecifically, we model the gripper-object interaction as a spring-mass system,\nwhere tactile sensors induce attractive forces, and proprioception generates\nrepulsive forces. We validate our framework through experiments on a real-world\nrobot setup, demonstrating its effectiveness across representative visual\nbackbones and manipulation scenarios, including grasping, object picking, and\nbimanual handover. Compared to the visual models, our approach overcomes some\ndrastic failure modes while tracking the in-hand object pose. In our\nexperiments, our approach shows an average increase of 55% in AUC of ADD-S and\n60% in ADD, along with an 80% lower position error compared to FoundationPose.\n", "date": "2025-04-17", "categories": []}
{"id": "2504.09160", "title": "SCFlow2: Plug-and-Play Object Pose Refiner with Shape-Constraint Scene\n  Flow", "abstract": "  We introduce SCFlow2, a plug-and-play refinement framework for 6D object pose\nestimation. Most recent 6D object pose methods rely on refinement to get\naccurate results. However, most existing refinement methods either suffer from\nnoises in establishing correspondences, or rely on retraining for novel\nobjects. SCFlow2 is based on the SCFlow model designed for refinement with\nshape constraint, but formulates the additional depth as a regularization in\nthe iteration via 3D scene flow for RGBD frames. The key design of SCFlow2 is\nan introduction of geometry constraints into the training of recurrent matching\nnetwork, by combining the rigid-motion embeddings in 3D scene flow and 3D shape\nprior of the target. We train SCFlow2 on a combination of dataset Objaverse,\nGSO and ShapeNet, and evaluate on BOP datasets with novel objects. After using\nour method as a post-processing, most state-of-the-art methods produce\nsignificantly better results, without any retraining or fine-tuning. The source\ncode is available at https://scflow2.github.io.\n", "date": "2025-04-12", "categories": []}
{"id": "2504.10433", "title": "MonoDiff9D: Monocular Category-Level 9D Object Pose Estimation via\n  Diffusion Model", "abstract": "  Object pose estimation is a core means for robots to understand and interact\nwith their environment. For this task, monocular category-level methods are\nattractive as they require only a single RGB camera. However, current methods\nrely on shape priors or CAD models of the intra-class known objects. We propose\na diffusion-based monocular category-level 9D object pose generation method,\nMonoDiff9D. Our motivation is to leverage the probabilistic nature of diffusion\nmodels to alleviate the need for shape priors, CAD models, or depth sensors for\nintra-class unknown object pose estimation. We first estimate coarse depth via\nDINOv2 from the monocular image in a zero-shot manner and convert it into a\npoint cloud. We then fuse the global features of the point cloud with the input\nimage and use the fused features along with the encoded time step to condition\nMonoDiff9D. Finally, we design a transformer-based denoiser to recover the\nobject pose from Gaussian noise. Extensive experiments on two popular benchmark\ndatasets show that MonoDiff9D achieves state-of-the-art monocular\ncategory-level 9D object pose estimation accuracy without the need for shape\npriors or CAD models at any stage. Our code will be made public at\nhttps://github.com/CNJianLiu/MonoDiff9D.\n", "date": "2025-04-14", "categories": []}
{"id": "2305.03535", "title": "Next-generation Surgical Navigation: Marker-less Multi-view 6DoF Pose\n  Estimation of Surgical Instruments", "abstract": "  State-of-the-art research of traditional computer vision is increasingly\nleveraged in the surgical domain. A particular focus in computer-assisted\nsurgery is to replace marker-based tracking systems for instrument localization\nwith pure image-based 6DoF pose estimation using deep-learning methods.\nHowever, state-of-the-art single-view pose estimation methods do not yet meet\nthe accuracy required for surgical navigation. In this context, we investigate\nthe benefits of multi-view setups for highly accurate and occlusion-robust 6DoF\npose estimation of surgical instruments and derive recommendations for an ideal\ncamera system that addresses the challenges in the operating room.\n  The contributions of this work are threefold. First, we present a\nmulti-camera capture setup consisting of static and head-mounted cameras, which\nallows us to study the performance of pose estimation methods under various\ncamera configurations. Second, we publish a multi-view RGB-D video dataset of\nex-vivo spine surgeries, captured in a surgical wet lab and a real operating\ntheatre and including rich annotations for surgeon, instrument, and patient\nanatomy. Third, we evaluate three state-of-the-art single-view and multi-view\nmethods for the task of 6DoF pose estimation of surgical instruments and\nanalyze the influence of camera configurations, training data, and occlusions\non the pose accuracy and generalization ability. The best method utilizes five\ncameras in a multi-view pose optimization and achieves an average position and\norientation error of 1.01 mm and 0.89{\\deg} for a surgical drill as well as\n2.79 mm and 3.33{\\deg} for a screwdriver under optimal conditions. Our results\ndemonstrate that marker-less tracking of surgical instruments is becoming a\nfeasible alternative to existing marker-based systems.\n", "date": "2023-05-05", "categories": []}
{"id": "2412.01543", "title": "6DOPE-GS: Online 6D Object Pose Estimation using Gaussian Splatting", "abstract": "  Efficient and accurate object pose estimation is an essential component for\nmodern vision systems in many applications such as Augmented Reality,\nautonomous driving, and robotics. While research in model-based 6D object pose\nestimation has delivered promising results, model-free methods are hindered by\nthe high computational load in rendering and inferring consistent poses of\narbitrary objects in a live RGB-D video stream. To address this issue, we\npresent 6DOPE-GS, a novel method for online 6D object pose estimation \\&\ntracking with a single RGB-D camera by effectively leveraging advances in\nGaussian Splatting. Thanks to the fast differentiable rendering capabilities of\nGaussian Splatting, 6DOPE-GS can simultaneously optimize for 6D object poses\nand 3D object reconstruction. To achieve the necessary efficiency and accuracy\nfor live tracking, our method uses incremental 2D Gaussian Splatting with an\nintelligent dynamic keyframe selection procedure to achieve high spatial object\ncoverage and prevent erroneous pose updates. We also propose an opacity\nstatistic-based pruning mechanism for adaptive Gaussian density control, to\nensure training stability and efficiency. We evaluate our method on the HO3D\nand YCBInEOAT datasets and show that 6DOPE-GS matches the performance of\nstate-of-the-art baselines for model-free simultaneous 6D pose tracking and\nreconstruction while providing a 5$\\times$ speedup. We also demonstrate the\nmethod's suitability for live, dynamic object tracking and reconstruction in a\nreal-world setting.\n", "date": "2024-12-02", "categories": []}
{"id": "2503.23587", "title": "PhysPose: Refining 6D Object Poses with Physical Constraints", "abstract": "  Accurate 6D object pose estimation from images is a key problem in\nobject-centric scene understanding, enabling applications in robotics,\naugmented reality, and scene reconstruction. Despite recent advances, existing\nmethods often produce physically inconsistent pose estimates, hindering their\ndeployment in real-world scenarios. We introduce PhysPose, a novel approach\nthat integrates physical reasoning into pose estimation through a\npostprocessing optimization enforcing non-penetration and gravitational\nconstraints. By leveraging scene geometry, PhysPose refines pose estimates to\nensure physical plausibility. Our approach achieves state-of-the-art accuracy\non the YCB-Video dataset from the BOP benchmark and improves over the\nstate-of-the-art pose estimation methods on the HOPE-Video dataset.\nFurthermore, we demonstrate its impact in robotics by significantly improving\nsuccess rates in a challenging pick-and-place task, highlighting the importance\nof physical consistency in real-world applications.\n", "date": "2025-03-30", "categories": []}
{"id": "2503.18673", "title": "Any6D: Model-free 6D Pose Estimation of Novel Objects", "abstract": "  We introduce Any6D, a model-free framework for 6D object pose estimation that\nrequires only a single RGB-D anchor image to estimate both the 6D pose and size\nof unknown objects in novel scenes. Unlike existing methods that rely on\ntextured 3D models or multiple viewpoints, Any6D leverages a joint object\nalignment process to enhance 2D-3D alignment and metric scale estimation for\nimproved pose accuracy. Our approach integrates a render-and-compare strategy\nto generate and refine pose hypotheses, enabling robust performance in\nscenarios with occlusions, non-overlapping views, diverse lighting conditions,\nand large cross-environment variations. We evaluate our method on five\nchallenging datasets: REAL275, Toyota-Light, HO3D, YCBINEOAT, and LM-O,\ndemonstrating its effectiveness in significantly outperforming state-of-the-art\nmethods for novel object pose estimation. Project page:\nhttps://taeyeop.com/any6d\n", "date": "2025-03-24", "categories": []}
{"id": "2503.17731", "title": "Co-op: Correspondence-based Novel Object Pose Estimation", "abstract": "  We propose Co-op, a novel method for accurately and robustly estimating the\n6DoF pose of objects unseen during training from a single RGB image. Our method\nrequires only the CAD model of the target object and can precisely estimate its\npose without any additional fine-tuning. While existing model-based methods\nsuffer from inefficiency due to using a large number of templates, our method\nenables fast and accurate estimation with a small number of templates. This\nimprovement is achieved by finding semi-dense correspondences between the input\nimage and the pre-rendered templates. Our method achieves strong generalization\nperformance by leveraging a hybrid representation that combines patch-level\nclassification and offset regression. Additionally, our pose refinement model\nestimates probabilistic flow between the input image and the rendered image,\nrefining the initial estimate to an accurate pose using a differentiable PnP\nlayer. We demonstrate that our method not only estimates object poses rapidly\nbut also outperforms existing methods by a large margin on the seven core\ndatasets of the BOP Challenge, achieving state-of-the-art accuracy.\n", "date": "2025-03-22", "categories": []}
{"id": "2409.18261", "title": "Omni6D: Large-Vocabulary 3D Object Dataset for Category-Level 6D Object\n  Pose Estimation", "abstract": "  6D object pose estimation aims at determining an object's translation,\nrotation, and scale, typically from a single RGBD image. Recent advancements\nhave expanded this estimation from instance-level to category-level, allowing\nmodels to generalize across unseen instances within the same category. However,\nthis generalization is limited by the narrow range of categories covered by\nexisting datasets, such as NOCS, which also tend to overlook common real-world\nchallenges like occlusion. To tackle these challenges, we introduce Omni6D, a\ncomprehensive RGBD dataset featuring a wide range of categories and varied\nbackgrounds, elevating the task to a more realistic context. 1) The dataset\ncomprises an extensive spectrum of 166 categories, 4688 instances adjusted to\nthe canonical pose, and over 0.8 million captures, significantly broadening the\nscope for evaluation. 2) We introduce a symmetry-aware metric and conduct\nsystematic benchmarks of existing algorithms on Omni6D, offering a thorough\nexploration of new challenges and insights. 3) Additionally, we propose an\neffective fine-tuning approach that adapts models from previous datasets to our\nextensive vocabulary setting. We believe this initiative will pave the way for\nnew insights and substantial progress in both the industrial and academic\nfields, pushing forward the boundaries of general 6D pose estimation.\n", "date": "2024-09-26", "categories": []}
{"id": "2503.15110", "title": "GIVEPose: Gradual Intra-class Variation Elimination for RGB-based\n  Category-Level Object Pose Estimation", "abstract": "  Recent advances in RGBD-based category-level object pose estimation have been\nlimited by their reliance on precise depth information, restricting their\nbroader applicability. In response, RGB-based methods have been developed.\nAmong these methods, geometry-guided pose regression that originated from\ninstance-level tasks has demonstrated strong performance. However, we argue\nthat the NOCS map is an inadequate intermediate representation for\ngeometry-guided pose regression method, as its many-to-one correspondence with\ncategory-level pose introduces redundant instance-specific information,\nresulting in suboptimal results. This paper identifies the intra-class\nvariation problem inherent in pose regression based solely on the NOCS map and\nproposes the Intra-class Variation-Free Consensus (IVFC) map, a novel\ncoordinate representation generated from the category-level consensus model. By\nleveraging the complementary strengths of the NOCS map and the IVFC map, we\nintroduce GIVEPose, a framework that implements Gradual Intra-class Variation\nElimination for category-level object pose estimation. Extensive evaluations on\nboth synthetic and real-world datasets demonstrate that GIVEPose significantly\noutperforms existing state-of-the-art RGB-based approaches, achieving\nsubstantial improvements in category-level object pose estimation. Our code is\navailable at https://github.com/ziqin-h/GIVEPose.\n", "date": "2025-03-19", "categories": []}
{"id": "2503.13926", "title": "Learning Shape-Independent Transformation via Spherical Representations\n  for Category-Level Object Pose Estimation", "abstract": "  Category-level object pose estimation aims to determine the pose and size of\nnovel objects in specific categories. Existing correspondence-based approaches\ntypically adopt point-based representations to establish the correspondences\nbetween primitive observed points and normalized object coordinates. However,\ndue to the inherent shape-dependence of canonical coordinates, these methods\nsuffer from semantic incoherence across diverse object shapes. To resolve this\nissue, we innovatively leverage the sphere as a shared proxy shape of objects\nto learn shape-independent transformation via spherical representations. Based\non this insight, we introduce a novel architecture called SpherePose, which\nyields precise correspondence prediction through three core designs. Firstly,\nWe endow the point-wise feature extraction with SO(3)-invariance, which\nfacilitates robust mapping between camera coordinate space and object\ncoordinate space regardless of rotation transformation. Secondly, the spherical\nattention mechanism is designed to propagate and integrate features among\nspherical anchors from a comprehensive perspective, thus mitigating the\ninterference of noise and incomplete point cloud. Lastly, a hyperbolic\ncorrespondence loss function is designed to distinguish subtle distinctions,\nwhich can promote the precision of correspondence prediction. Experimental\nresults on CAMERA25, REAL275 and HouseCat6D benchmarks demonstrate the superior\nperformance of our method, verifying the effectiveness of spherical\nrepresentations and architectural innovations.\n", "date": "2025-03-18", "categories": []}
{"id": "2403.13683", "title": "DVMNet++: Rethinking Relative Pose Estimation for Unseen Objects", "abstract": "  Determining the relative pose of a previously unseen object between two\nimages is pivotal to the success of generalizable object pose estimation.\nExisting approaches typically predict 3D translation utilizing the ground-truth\nobject bounding box and approximate 3D rotation with a large number of discrete\nhypotheses. This strategy makes unrealistic assumptions about the availability\nof ground truth and incurs a computationally expensive process of scoring each\nhypothesis at test time. By contrast, we rethink the problem of relative pose\nestimation for unseen objects by presenting a Deep Voxel Matching Network\n(DVMNet++). Our method computes the relative object pose in a single pass,\neliminating the need for ground-truth object bounding boxes and rotation\nhypotheses. We achieve open-set object detection by leveraging image feature\nembedding and natural language understanding as reference. The detection result\nis then employed to approximate the translation parameters and crop the object\nfrom the query image. For rotation estimation, we map the two RGB images, i.e.,\nreference and cropped query, to their respective voxelized 3D representations.\nThe resulting voxels are passed through a rotation estimation module, which\naligns the voxels and computes the rotation in an end-to-end fashion by solving\na least-squares problem. To enhance robustness, we introduce a weighted closest\nvoxel algorithm capable of mitigating the impact of noisy voxels. We conduct\nextensive experiments on the CO3D, Objaverse, LINEMOD, and LINEMOD-O datasets,\ndemonstrating that our approach delivers more accurate relative pose estimates\nfor novel objects at a lower computational cost compared to state-of-the-art\nmethods. Our code is released at https://github.com/sailor-z/DVMNet/.\n", "date": "2024-03-20", "categories": []}
{"id": "2411.16106", "title": "UNOPose: Unseen Object Pose Estimation with an Unposed RGB-D Reference\n  Image", "abstract": "  Unseen object pose estimation methods often rely on CAD models or multiple\nreference views, making the onboarding stage costly. To simplify reference\nacquisition, we aim to estimate the unseen object's pose through a single\nunposed RGB-D reference image. While previous works leverage reference images\nas pose anchors to limit the range of relative pose, our scenario presents\nsignificant challenges since the relative transformation could vary across the\nentire SE(3) space. Moreover, factors like occlusion, sensor noise, and extreme\ngeometry could result in low viewpoint overlap. To address these challenges, we\npresent a novel approach and benchmark, termed UNOPose, for unseen\none-reference-based object pose estimation. Building upon a coarse-to-fine\nparadigm, UNOPose constructs an SE(3)-invariant reference frame to standardize\nobject representation despite pose and size variations. To alleviate small\noverlap across viewpoints, we recalibrate the weight of each correspondence\nbased on its predicted likelihood of being within the overlapping region.\nEvaluated on our proposed benchmark based on the BOP Challenge, UNOPose\ndemonstrates superior performance, significantly outperforming traditional and\nlearning-based methods in the one-reference setting and remaining competitive\nwith CAD-model-based methods. The code and dataset are available at\nhttps://github.com/shanice-l/UNOPose.\n", "date": "2024-11-25", "categories": []}
{"id": "2502.02525", "title": "Diff9D: Diffusion-Based Domain-Generalized Category-Level 9-DoF Object\n  Pose Estimation", "abstract": "  Nine-degrees-of-freedom (9-DoF) object pose and size estimation is crucial\nfor enabling augmented reality and robotic manipulation. Category-level methods\nhave received extensive research attention due to their potential for\ngeneralization to intra-class unknown objects. However, these methods require\nmanual collection and labeling of large-scale real-world training data. To\naddress this problem, we introduce a diffusion-based paradigm for\ndomain-generalized category-level 9-DoF object pose estimation. Our motivation\nis to leverage the latent generalization ability of the diffusion model to\naddress the domain generalization challenge in object pose estimation. This\nentails training the model exclusively on rendered synthetic data to achieve\ngeneralization to real-world scenes. We propose an effective diffusion model to\nredefine 9-DoF object pose estimation from a generative perspective. Our model\ndoes not require any 3D shape priors during training or inference. By employing\nthe Denoising Diffusion Implicit Model, we demonstrate that the reverse\ndiffusion process can be executed in as few as 3 steps, achieving near\nreal-time performance. Finally, we design a robotic grasping system comprising\nboth hardware and software components. Through comprehensive experiments on two\nbenchmark datasets and the real-world robotic system, we show that our method\nachieves state-of-the-art domain generalization performance. Our code will be\nmade public at https://github.com/CNJianLiu/Diff9D.\n", "date": "2025-02-04", "categories": []}
{"id": "2503.12732", "title": "Stereo Event-based, 6-DOF Pose Tracking for Uncooperative Spacecraft", "abstract": "  Pose tracking of uncooperative spacecraft is an essential technology for\nspace exploration and on-orbit servicing, which remains an open problem. Event\ncameras possess numerous advantages, such as high dynamic range, high temporal\nresolution, and low power consumption. These attributes hold the promise of\novercoming challenges encountered by conventional cameras, including motion\nblur and extreme illumination, among others. To address the standard on-orbit\nobservation missions, we propose a line-based pose tracking method for\nuncooperative spacecraft utilizing a stereo event camera. To begin with, we\nestimate the wireframe model of uncooperative spacecraft, leveraging the\nspatio-temporal consistency of stereo event streams for line-based\nreconstruction. Then, we develop an effective strategy to establish\ncorrespondences between events and projected lines of uncooperative spacecraft.\nUsing these correspondences, we formulate the pose tracking as a continuous\noptimization process over 6-DOF motion parameters, achieved by minimizing\nevent-line distances. Moreover, we construct a stereo event-based uncooperative\nspacecraft motion dataset, encompassing both simulated and real events. The\nproposed method is quantitatively evaluated through experiments conducted on\nour self-collected dataset, demonstrating an improvement in terms of\neffectiveness and accuracy over competing methods. The code will be\nopen-sourced at https://github.com/Zibin6/SE6PT.\n", "date": "2025-03-17", "categories": []}
{"id": "2503.10307", "title": "6D Object Pose Tracking in Internet Videos for Robotic Manipulation", "abstract": "  We seek to extract a temporally consistent 6D pose trajectory of a\nmanipulated object from an Internet instructional video. This is a challenging\nset-up for current 6D pose estimation methods due to uncontrolled capturing\nconditions, subtle but dynamic object motions, and the fact that the exact mesh\nof the manipulated object is not known. To address these challenges, we present\nthe following contributions. First, we develop a new method that estimates the\n6D pose of any object in the input image without prior knowledge of the object\nitself. The method proceeds by (i) retrieving a CAD model similar to the\ndepicted object from a large-scale model database, (ii) 6D aligning the\nretrieved CAD model with the input image, and (iii) grounding the absolute\nscale of the object with respect to the scene. Second, we extract smooth 6D\nobject trajectories from Internet videos by carefully tracking the detected\nobjects across video frames. The extracted object trajectories are then\nretargeted via trajectory optimization into the configuration space of a\nrobotic manipulator. Third, we thoroughly evaluate and ablate our 6D pose\nestimation method on YCB-V and HOPE-Video datasets as well as a new dataset of\ninstructional videos manually annotated with approximate 6D object\ntrajectories. We demonstrate significant improvements over existing\nstate-of-the-art RGB 6D pose estimation methods. Finally, we show that the 6D\nobject motion estimated from Internet videos can be transferred to a 7-axis\nrobotic manipulator both in a virtual simulator as well as in a real world\nset-up. We also successfully apply our method to egocentric videos taken from\nthe EPIC-KITCHENS dataset, demonstrating potential for Embodied AI\napplications.\n", "date": "2025-03-13", "categories": []}
{"id": "2502.13403", "title": "Object-Pose Estimation With Neural Population Codes", "abstract": "  Robotic assembly tasks require object-pose estimation, particularly for tasks\nthat avoid costly mechanical constraints. Object symmetry complicates the\ndirect mapping of sensory input to object rotation, as the rotation becomes\nambiguous and lacks a unique training target. Some proposed solutions involve\nevaluating multiple pose hypotheses against the input or predicting a\nprobability distribution, but these approaches suffer from significant\ncomputational overhead. Here, we show that representing object rotation with a\nneural population code overcomes these limitations, enabling a direct mapping\nto rotation and end-to-end learning. As a result, population codes facilitate\nfast and accurate pose estimation. On the T-LESS dataset, we achieve inference\nin 3.2 milliseconds on an Apple M1 CPU and a Maximum Symmetry-Aware Surface\nDistance accuracy of 84.7% using only gray-scale image input, compared to 69.7%\naccuracy when directly mapping to pose.\n", "date": "2025-02-19", "categories": []}
{"id": "2503.06660", "title": "AxisPose: Model-Free Matching-Free Single-Shot 6D Object Pose Estimation\n  via Axis Generation", "abstract": "  Object pose estimation, which plays a vital role in robotics, augmented\nreality, and autonomous driving, has been of great interest in computer vision.\nExisting studies either require multi-stage pose regression or rely on 2D-3D\nfeature matching. Though these approaches have shown promising results, they\nrely heavily on appearance information, requiring complex input (i.e.,\nmulti-view reference input, depth, or CAD models) and intricate pipeline (i.e.,\nfeature extraction-SfM-2D to 3D matching-PnP). We propose AxisPose, a\nmodel-free, matching-free, single-shot solution for robust 6D pose estimation,\nwhich fundamentally diverges from the existing paradigm. Unlike existing\nmethods that rely on 2D-3D or 2D-2D matching using 3D techniques, such as SfM\nand PnP, AxisPose directly infers a robust 6D pose from a single view by\nleveraging a diffusion model to learn the latent axis distribution of objects\nwithout reference views. Specifically, AxisPose constructs an Axis Generation\nModule (AGM) to capture the latent geometric distribution of object axes\nthrough a diffusion model. The diffusion process is guided by injecting the\ngradient of geometric consistency loss into the noise estimation to maintain\nthe geometric consistency of the generated tri-axis. With the generated\ntri-axis projection, AxisPose further adopts a Triaxial Back-projection Module\n(TBM) to recover the 6D pose from the object tri-axis. The proposed AxisPose\nachieves robust performance at the cross-instance level (i.e., one model for N\ninstances) using only a single view as input without reference images, with\ngreat potential for generalization to unseen-object level.\n", "date": "2025-03-09", "categories": []}
{"id": "2501.18478", "title": "SimpleDepthPose: Fast and Reliable Human Pose Estimation with\n  RGBD-Images", "abstract": "  In the rapidly advancing domain of computer vision, accurately estimating the\nposes of multiple individuals from various viewpoints remains a significant\nchallenge, especially when reliability is a key requirement. This paper\nintroduces a novel algorithm that excels in multi-view, multi-person pose\nestimation by incorporating depth information. An extensive evaluation\ndemonstrates that the proposed algorithm not only generalizes well to unseen\ndatasets, and shows a fast runtime performance, but also is adaptable to\ndifferent keypoints. To support further research, all of the work is publicly\naccessible.\n", "date": "2025-01-30", "categories": []}
{"id": "2503.05174", "title": "SplatPose: Geometry-Aware 6-DoF Pose Estimation from Single RGB Image\n  via 3D Gaussian Splatting", "abstract": "  6-DoF pose estimation is a fundamental task in computer vision with\nwide-ranging applications in augmented reality and robotics. Existing single\nRGB-based methods often compromise accuracy due to their reliance on initial\npose estimates and susceptibility to rotational ambiguity, while approaches\nrequiring depth sensors or multi-view setups incur significant deployment\ncosts. To address these limitations, we introduce SplatPose, a novel framework\nthat synergizes 3D Gaussian Splatting (3DGS) with a dual-branch neural\narchitecture to achieve high-precision pose estimation using only a single RGB\nimage. Central to our approach is the Dual-Attention Ray Scoring Network\n(DARS-Net), which innovatively decouples positional and angular alignment\nthrough geometry-domain attention mechanisms, explicitly modeling directional\ndependencies to mitigate rotational ambiguity. Additionally, a coarse-to-fine\noptimization pipeline progressively refines pose estimates by aligning dense 2D\nfeatures between query images and 3DGS-synthesized views, effectively\ncorrecting feature misalignment and depth errors from sparse ray sampling.\nExperiments on three benchmark datasets demonstrate that SplatPose achieves\nstate-of-the-art 6-DoF pose estimation accuracy in single RGB settings,\nrivaling approaches that depend on depth or multi-view images.\n", "date": "2025-03-07", "categories": []}
{"id": "2503.03655", "title": "Improving 6D Object Pose Estimation of metallic Household and Industry\n  Objects", "abstract": "  6D object pose estimation suffers from reduced accuracy when applied to\nmetallic objects. We set out to improve the state-of-the-art by addressing\nchallenges such as reflections and specular highlights in industrial\napplications. Our novel BOP-compatible dataset, featuring a diverse set of\nmetallic objects (cans, household, and industrial items) under various lighting\nand background conditions, provides additional geometric and visual cues. We\ndemonstrate that these cues can be effectively leveraged to enhance overall\nperformance. To illustrate the usefulness of the additional features, we\nimprove upon the GDRNPP algorithm by introducing an additional keypoint\nprediction and material estimator head in order to improve spatial scene\nunderstanding. Evaluations on the new dataset show improved accuracy for\nmetallic objects, supporting the hypothesis that additional geometric and\nvisual cues can improve learning.\n", "date": "2025-03-05", "categories": []}
{"id": "2502.16419", "title": "DeProPose: Deficiency-Proof 3D Human Pose Estimation via Adaptive\n  Multi-View Fusion", "abstract": "  3D human pose estimation has wide applications in fields such as intelligent\nsurveillance, motion capture, and virtual reality. However, in real-world\nscenarios, issues such as occlusion, noise interference, and missing viewpoints\ncan severely affect pose estimation. To address these challenges, we introduce\nthe task of Deficiency-Aware 3D Pose Estimation. Traditional 3D pose estimation\nmethods often rely on multi-stage networks and modular combinations, which can\nlead to cumulative errors and increased training complexity, making them unable\nto effectively address deficiency-aware estimation. To this end, we propose\nDeProPose, a flexible method that simplifies the network architecture to reduce\ntraining complexity and avoid information loss in multi-stage designs.\nAdditionally, the model innovatively introduces a multi-view feature fusion\nmechanism based on relative projection error, which effectively utilizes\ninformation from multiple viewpoints and dynamically assigns weights, enabling\nefficient integration and enhanced robustness to overcome deficiency-aware 3D\nPose Estimation challenges. Furthermore, to thoroughly evaluate this end-to-end\nmulti-view 3D human pose estimation model and to advance research on\nocclusion-related challenges, we have developed a novel 3D human pose\nestimation dataset, termed the Deficiency-Aware 3D Pose Estimation (DA-3DPE)\ndataset. This dataset encompasses a wide range of deficiency scenarios,\nincluding noise interference, missing viewpoints, and occlusion challenges.\nCompared to state-of-the-art methods, DeProPose not only excels in addressing\nthe deficiency-aware problem but also shows improvement in conventional\nscenarios, providing a powerful and user-friendly solution for 3D human pose\nestimation. The source code will be available at\nhttps://github.com/WUJINHUAN/DeProPose.\n", "date": "2025-02-23", "categories": []}
{"id": "2502.14061", "title": "EfficientPose 6D: Scalable and Efficient 6D Object Pose Estimation", "abstract": "  In industrial applications requiring real-time feedback, such as quality\ncontrol and robotic manipulation, the demand for high-speed and accurate pose\nestimation remains critical. Despite advances improving speed and accuracy in\npose estimation, finding a balance between computational efficiency and\naccuracy poses significant challenges in dynamic environments. Most current\nalgorithms lack scalability in estimation time, especially for diverse\ndatasets, and the state-of-the-art (SOTA) methods are often too slow. This\nstudy focuses on developing a fast and scalable set of pose estimators based on\nGDRNPP to meet or exceed current benchmarks in accuracy and robustness,\nparticularly addressing the efficiency-accuracy trade-off essential in\nreal-time scenarios. We propose the AMIS algorithm to tailor the utilized model\naccording to an application-specific trade-off between inference time and\naccuracy. We further show the effectiveness of the AMIS-based model choice on\nfour prominent benchmark datasets (LM-O, YCB-V, T-LESS, and ITODD).\n", "date": "2025-02-19", "categories": []}
{"id": "2502.10606", "title": "HIPPo: Harnessing Image-to-3D Priors for Model-free Zero-shot 6D Pose\n  Estimation", "abstract": "  This work focuses on model-free zero-shot 6D object pose estimation for\nrobotics applications. While existing methods can estimate the precise 6D pose\nof objects, they heavily rely on curated CAD models or reference images, the\npreparation of which is a time-consuming and labor-intensive process. Moreover,\nin real-world scenarios, 3D models or reference images may not be available in\nadvance and instant robot reaction is desired. In this work, we propose a novel\nframework named HIPPo, which eliminates the need for curated CAD models and\nreference images by harnessing image-to-3D priors from Diffusion Models,\nenabling model-free zero-shot 6D pose estimation. Specifically, we construct\nHIPPo Dreamer, a rapid image-to-mesh model built on a multiview Diffusion Model\nand a 3D reconstruction foundation model. Our HIPPo Dreamer can generate a 3D\nmesh of any unseen objects from a single glance in just a few seconds. Then, as\nmore observations are acquired, we propose to continuously refine the diffusion\nprior mesh model by joint optimization of object geometry and appearance. This\nis achieved by a measurement-guided scheme that gradually replaces the\nplausible diffusion priors with more reliable online observations.\nConsequently, HIPPo can instantly estimate and track the 6D pose of a novel\nobject and maintain a complete mesh for immediate robotic applications.\nThorough experiments on various benchmarks show that HIPPo outperforms\nstate-of-the-art methods in 6D object pose estimation when prior reference\nimages are limited.\n", "date": "2025-02-14", "categories": []}
{"id": "2502.12027", "title": "Enhancing Transparent Object Pose Estimation: A Fusion of GDR-Net and\n  Edge Detection", "abstract": "  Object pose estimation of transparent objects remains a challenging task in\nthe field of robot vision due to the immense influence of lighting, background,\nand reflections. However, the edges of clear objects have the highest contrast,\nwhich leads to stable and prominent features. We propose a novel approach by\nincorporating edge detection in a pre-processing step for the tasks of object\ndetection and object pose estimation. We conducted experiments to investigate\nthe effect of edge detectors on transparent objects. We examine the performance\nof the state-of-the-art 6D object pose estimation pipeline GDR-Net and the\nobject detector YOLOX when applying different edge detectors as pre-processing\nsteps (i.e., Canny edge detection with and without color information, and\nholistically-nested edges (HED)). We evaluate the physically-based rendered\ndataset Trans6D-32 K of transparent objects with parameters proposed by the BOP\nChallenge. Our results indicate that applying edge detection as a\npre-processing enhances performance for certain objects.\n", "date": "2025-02-17", "categories": []}
{"id": "2502.03877", "title": "Advanced Object Detection and Pose Estimation with Hybrid Task Cascade\n  and High-Resolution Networks", "abstract": "  In the field of computer vision, 6D object detection and pose estimation are\ncritical for applications such as robotics, augmented reality, and autonomous\ndriving. Traditional methods often struggle with achieving high accuracy in\nboth object detection and precise pose estimation simultaneously. This study\nproposes an improved 6D object detection and pose estimation pipeline based on\nthe existing 6D-VNet framework, enhanced by integrating a Hybrid Task Cascade\n(HTC) and a High-Resolution Network (HRNet) backbone. By leveraging the\nstrengths of HTC's multi-stage refinement process and HRNet's ability to\nmaintain high-resolution representations, our approach significantly improves\ndetection accuracy and pose estimation precision. Furthermore, we introduce\nadvanced post-processing techniques and a novel model integration strategy that\ncollectively contribute to superior performance on public and private\nbenchmarks. Our method demonstrates substantial improvements over\nstate-of-the-art models, making it a valuable contribution to the domain of 6D\nobject detection and pose estimation.\n", "date": "2025-02-06", "categories": []}
{"id": "2502.01004", "title": "ZeroBP: Learning Position-Aware Correspondence for Zero-shot 6D Pose\n  Estimation in Bin-Picking", "abstract": "  Bin-picking is a practical and challenging robotic manipulation task, where\naccurate 6D pose estimation plays a pivotal role. The workpieces in bin-picking\nare typically textureless and randomly stacked in a bin, which poses a\nsignificant challenge to 6D pose estimation. Existing solutions are typically\nlearning-based methods, which require object-specific training. Their\nefficiency of practical deployment for novel workpieces is highly limited by\ndata collection and model retraining. Zero-shot 6D pose estimation is a\npotential approach to address the issue of deployment efficiency. Nevertheless,\nexisting zero-shot 6D pose estimation methods are designed to leverage feature\nmatching to establish point-to-point correspondences for pose estimation, which\nis less effective for workpieces with textureless appearances and ambiguous\nlocal regions. In this paper, we propose ZeroBP, a zero-shot pose estimation\nframework designed specifically for the bin-picking task. ZeroBP learns\nPosition-Aware Correspondence (PAC) between the scene instance and its CAD\nmodel, leveraging both local features and global positions to resolve the\nmismatch issue caused by ambiguous regions with similar shapes and appearances.\nExtensive experiments on the ROBI dataset demonstrate that ZeroBP outperforms\nstate-of-the-art zero-shot pose estimation methods, achieving an improvement of\n9.1% in average recall of correct poses.\n", "date": "2025-02-03", "categories": []}
{"id": "2501.14896", "title": "Glissando-Net: Deep sinGLe vIew category level poSe eStimation ANd 3D\n  recOnstruction", "abstract": "  We present a deep learning model, dubbed Glissando-Net, to simultaneously\nestimate the pose and reconstruct the 3D shape of objects at the category level\nfrom a single RGB image. Previous works predominantly focused on either\nestimating poses(often at the instance level), or reconstructing shapes, but\nnot both. Glissando-Net is composed of two auto-encoders that are jointly\ntrained, one for RGB images and the other for point clouds. We embrace two key\ndesign choices in Glissando-Net to achieve a more accurate prediction of the 3D\nshape and pose of the object given a single RGB image as input. First, we\naugment the feature maps of the point cloud encoder and decoder with\ntransformed feature maps from the image decoder, enabling effective 2D-3D\ninteraction in both training and prediction. Second, we predict both the 3D\nshape and pose of the object in the decoder stage. This way, we better utilize\nthe information in the 3D point clouds presented only in the training stage to\ntrain the network for more accurate prediction. We jointly train the two\nencoder-decoders for RGB and point cloud data to learn how to pass latent\nfeatures to the point cloud decoder during inference. In testing, the encoder\nof the 3D point cloud is discarded. The design of Glissando-Net is inspired by\ncodeSLAM. Unlike codeSLAM, which targets 3D reconstruction of scenes, we focus\non pose estimation and shape reconstruction of objects, and directly predict\nthe object pose and a pose invariant 3D reconstruction without the need of the\ncode optimization step. Extensive experiments, involving both ablation studies\nand comparison with competing methods, demonstrate the efficacy of our proposed\nmethod, and compare favorably with the state-of-the-art.\n", "date": "2025-01-24", "categories": []}
{"id": "2501.13009", "title": "Deep Learning-Based Image Recovery and Pose Estimation for Resident\n  Space Objects", "abstract": "  As the density of spacecraft in Earth's orbit increases, their recognition,\npose and trajectory identification becomes crucial for averting potential\ncollisions and executing debris removal operations. However, training models\nable to identify a spacecraft and its pose presents a significant challenge due\nto a lack of available image data for model training. This paper puts forth an\ninnovative framework for generating realistic synthetic datasets of Resident\nSpace Object (RSO) imagery. Using the International Space Station (ISS) as a\ntest case, it goes on to combine image regression with image restoration\nmethodologies to estimate pose from blurred images. An analysis of the proposed\nimage recovery and regression techniques was undertaken, providing insights\ninto the performance, potential enhancements and limitations when applied to\nreal imagery of RSOs. The image recovery approach investigated involves first\napplying image deconvolution using an effective point spread function, followed\nby detail object extraction with a U-Net. Interestingly, using only U-Net for\nimage reconstruction the best pose performance was attained, reducing the\naverage Mean Squared Error in image recovery by 97.28% and the average angular\nerror by 71.9%. The successful application of U-Net image restoration combined\nwith the Resnet50 regression network for pose estimation of the International\nSpace Station demonstrates the value of a diverse set of evaluation tools for\neffective solutions to real-world problems such as the analysis of distant\nobjects in Earth's orbit.\n", "date": "2025-01-22", "categories": []}
{"id": "2501.09565", "title": "A New Teacher-Reviewer-Student Framework for Semi-supervised 2D Human\n  Pose Estimation", "abstract": "  Conventional 2D human pose estimation methods typically require extensive\nlabeled annotations, which are both labor-intensive and expensive. In contrast,\nsemi-supervised 2D human pose estimation can alleviate the above problems by\nleveraging a large amount of unlabeled data along with a small portion of\nlabeled data. Existing semi-supervised 2D human pose estimation methods update\nthe network through backpropagation, ignoring crucial historical information\nfrom the previous training process. Therefore, we propose a novel\nsemi-supervised 2D human pose estimation method by utilizing a newly designed\nTeacher-Reviewer-Student framework. Specifically, we first mimic the phenomenon\nthat human beings constantly review previous knowledge for consolidation to\ndesign our framework, in which the teacher predicts results to guide the\nstudent's learning and the reviewer stores important historical parameters to\nprovide additional supervision signals. Secondly, we introduce a Multi-level\nFeature Learning strategy, which utilizes the outputs from different stages of\nthe backbone to estimate the heatmap to guide network training, enriching the\nsupervisory information while effectively capturing keypoint relationships.\nFinally, we design a data augmentation strategy, i.e., Keypoint-Mix, to perturb\npose information by mixing different keypoints, thus enhancing the network's\nability to discern keypoints. Extensive experiments on publicly available\ndatasets, demonstrate our method achieves significant improvements compared to\nthe existing methods.\n", "date": "2025-01-16", "categories": []}
{"id": "2312.00947", "title": "FreeZe: Training-free zero-shot 6D pose estimation with geometric and\n  vision foundation models", "abstract": "  Estimating the 6D pose of objects unseen during training is highly desirable\nyet challenging. Zero-shot object 6D pose estimation methods address this\nchallenge by leveraging additional task-specific supervision provided by\nlarge-scale, photo-realistic synthetic datasets. However, their performance\nheavily depends on the quality and diversity of rendered data and they require\nextensive training. In this work, we show how to tackle the same task but\nwithout training on specific data. We propose FreeZe, a novel solution that\nharnesses the capabilities of pre-trained geometric and vision foundation\nmodels. FreeZe leverages 3D geometric descriptors learned from unrelated 3D\npoint clouds and 2D visual features learned from web-scale 2D images to\ngenerate discriminative 3D point-level descriptors. We then estimate the 6D\npose of unseen objects by 3D registration based on RANSAC. We also introduce a\nnovel algorithm to solve ambiguous cases due to geometrically symmetric objects\nthat is based on visual features. We comprehensively evaluate FreeZe across the\nseven core datasets of the BOP Benchmark, which include over a hundred 3D\nobjects and 20,000 images captured in various scenarios. FreeZe consistently\noutperforms all state-of-the-art approaches, including competitors extensively\ntrained on synthetic 6D pose estimation data. Code will be publicly available\nat https://andreacaraffa.github.io/freeze.\n", "date": "2023-12-01", "categories": []}
{"id": "2501.02831", "title": "Universal Features Guided Zero-Shot Category-Level Object Pose\n  Estimation", "abstract": "  Object pose estimation, crucial in computer vision and robotics applications,\nfaces challenges with the diversity of unseen categories. We propose a\nzero-shot method to achieve category-level 6-DOF object pose estimation, which\nexploits both 2D and 3D universal features of input RGB-D image to establish\nsemantic similarity-based correspondences and can be extended to unseen\ncategories without additional model fine-tuning. Our method begins with\ncombining efficient 2D universal features to find sparse correspondences\nbetween intra-category objects and gets initial coarse pose. To handle the\ncorrespondence degradation of 2D universal features if the pose deviates much\nfrom the target pose, we use an iterative strategy to optimize the pose.\nSubsequently, to resolve pose ambiguities due to shape differences between\nintra-category objects, the coarse pose is refined by optimizing with dense\nalignment constraint of 3D universal features. Our method outperforms previous\nmethods on the REAL275 and Wild6D benchmarks for unseen categories.\n", "date": "2025-01-06", "categories": []}
{"id": "2501.02916", "title": "Spiking monocular event based 6D pose estimation for space application", "abstract": "  With the growing interest in on On-orbit servicing (OOS) and Active Debris\nRemoval (ADR) missions, spacecraft poses estimation algorithms are being\ndeveloped using deep learning to improve the precision of this complex task and\nfind the most efficient solution. With the advances of bio-inspired low-power\nsolutions, such a spiking neural networks and event-based processing and\ncameras, and their recent work for space applications, we propose to\ninvestigate the feasibility of a fully event-based solution to improve\nevent-based pose estimation for spacecraft. In this paper, we address the first\nevent-based dataset SEENIC with real event frames captured by an event-based\ncamera on a testbed. We show the methods and results of the first event-based\nsolution for this use case, where our small spiking end-to-end network (S2E2)\nsolution achieves interesting results over 21cm position error and 14degree\nrotation error, which is the first step towards fully event-based processing\nfor embedded spacecraft pose estimation.\n", "date": "2025-01-06", "categories": []}
{"id": "2412.20830", "title": "ReFlow6D: Refraction-Guided Transparent Object 6D Pose Estimation via\n  Intermediate Representation Learning", "abstract": "  Transparent objects are ubiquitous in daily life, making their perception and\nrobotics manipulation important. However, they present a major challenge due to\ntheir distinct refractive and reflective properties when it comes to accurately\nestimating the 6D pose. To solve this, we present ReFlow6D, a novel method for\ntransparent object 6D pose estimation that harnesses the\nrefractive-intermediate representation. Unlike conventional approaches, our\nmethod leverages a feature space impervious to changes in RGB image space and\nindependent of depth information. Drawing inspiration from image matting, we\nmodel the deformation of the light path through transparent objects, yielding a\nunique object-specific intermediate representation guided by light refraction\nthat is independent of the environment in which objects are observed. By\nintegrating these intermediate features into the pose estimation network, we\nshow that ReFlow6D achieves precise 6D pose estimation of transparent objects,\nusing only RGB images as input. Our method further introduces a novel\ntransparent object compositing loss, fostering the generation of superior\nrefractive-intermediate features. Empirical evaluations show that our approach\nsignificantly outperforms state-of-the-art methods on TOD and Trans32K-6D\ndatasets. Robot grasping experiments further demonstrate that ReFlow6D's pose\nestimation accuracy effectively translates to real-world robotics task. The\nsource code is available at: https://github.com/StoicGilgamesh/ReFlow6D and\nhttps://github.com/StoicGilgamesh/matting_rendering.\n", "date": "2024-12-30", "categories": []}
{"id": "2412.18605", "title": "Orient Anything: Learning Robust Object Orientation Estimation from\n  Rendering 3D Models", "abstract": "  Orientation is a key attribute of objects, crucial for understanding their\nspatial pose and arrangement in images. However, practical solutions for\naccurate orientation estimation from a single image remain underexplored. In\nthis work, we introduce Orient Anything, the first expert and foundational\nmodel designed to estimate object orientation in a single- and free-view image.\nDue to the scarcity of labeled data, we propose extracting knowledge from the\n3D world. By developing a pipeline to annotate the front face of 3D objects and\nrender images from random views, we collect 2M images with precise orientation\nannotations. To fully leverage the dataset, we design a robust training\nobjective that models the 3D orientation as probability distributions of three\nangles and predicts the object orientation by fitting these distributions.\nBesides, we employ several strategies to improve synthetic-to-real transfer.\nOur model achieves state-of-the-art orientation estimation accuracy in both\nrendered and real images and exhibits impressive zero-shot ability in various\nscenarios. More importantly, our model enhances many applications, such as\ncomprehension and generation of complex spatial concepts and 3D object pose\nadjustment.\n", "date": "2024-12-24", "categories": []}
{"id": "2211.11188", "title": "Simultaneous Multiple Object Detection and Pose Estimation using 3D\n  Model Infusion with Monocular Vision", "abstract": "  Multiple object detection and pose estimation are vital computer vision\ntasks. The latter relates to the former as a downstream problem in applications\nsuch as robotics and autonomous driving. However, due to the high complexity of\nboth tasks, existing methods generally treat them independently, which is\nsub-optimal. We propose simultaneous neural modeling of both using monocular\nvision and 3D model infusion. Our Simultaneous Multiple Object detection and\nPose Estimation network (SMOPE-Net) is an end-to-end trainable multitasking\nnetwork with a composite loss that also provides the advantages of anchor-free\ndetections for efficient downstream pose estimation. To enable the annotation\nof training data for our learning objective, we develop a Twin-Space object\nlabeling method and demonstrate its correctness analytically and empirically.\nUsing the labeling method, we provide the KITTI-6DoF dataset with $\\sim7.5$K\nannotated frames. Extensive experiments on KITTI-6DoF and the popular LineMod\ndatasets show a consistent performance gain with SMOPE-Net over existing pose\nestimation methods. Here are links to our proposed SMOPE-Net, KITTI-6DoF\ndataset, and LabelImg3D labeling tool.\n", "date": "2022-11-21", "categories": []}
{"id": "2412.16155", "title": "Can Generative Video Models Help Pose Estimation?", "abstract": "  Pairwise pose estimation from images with little or no overlap is an open\nchallenge in computer vision. Existing methods, even those trained on\nlarge-scale datasets, struggle in these scenarios due to the lack of\nidentifiable correspondences or visual overlap. Inspired by the human ability\nto infer spatial relationships from diverse scenes, we propose a novel\napproach, InterPose, that leverages the rich priors encoded within pre-trained\ngenerative video models. We propose to use a video model to hallucinate\nintermediate frames between two input images, effectively creating a dense,\nvisual transition, which significantly simplifies the problem of pose\nestimation. Since current video models can still produce implausible motion or\ninconsistent geometry, we introduce a self-consistency score that evaluates the\nconsistency of pose predictions from sampled videos. We demonstrate that our\napproach generalizes among three state-of-the-art video models and show\nconsistent improvements over the state-of-the-art DUSt3R on four diverse\ndatasets encompassing indoor, outdoor, and object-centric scenes. Our findings\nsuggest a promising avenue for improving pose estimation models by leveraging\nlarge generative models trained on vast amounts of video data, which is more\nreadily available than 3D data. See our project page for results:\nhttps://inter-pose.github.io/.\n", "date": "2024-12-20", "categories": []}
{"id": "2404.16471", "title": "COBRA -- COnfidence score Based on shape Regression Analysis for\n  method-independent quality assessment of object pose estimation from single\n  images", "abstract": "  We propose a generic procedure for assessing 6D object pose estimates. Our\napproach relies on the evaluation of discrepancies in the geometry of the\nobserved object, in particular its respective estimated back-projection in 3D,\nagainst a putative functional shape representation comprising mixtures of\nGaussian Processes, that act as a template. Each Gaussian Process is trained to\nyield a fragment of the object's surface in a radial fashion with respect to\ndesignated reference points. We further define a pose confidence measure as the\naverage probability of pixel back-projections in the Gaussian mixture. The goal\nof our experiments is two-fold. a) We demonstrate that our functional\nrepresentation is sufficiently accurate as a shape template on which the\nprobability of back-projected object points can be evaluated, and, b) we show\nthat the resulting confidence scores based on these probabilities are indeed a\nconsistent quality measure of pose.\n", "date": "2024-04-25", "categories": []}
{"id": "2412.11420", "title": "Category Level 6D Object Pose Estimation from a Single RGB Image using\n  Diffusion", "abstract": "  Estimating the 6D pose and 3D size of an object from an image is a\nfundamental task in computer vision. Most current approaches are restricted to\nspecific instances with known models or require ground truth depth information\nor point cloud captures from LIDAR. We tackle the harder problem of pose\nestimation for category-level objects from a single RGB image. We propose a\nnovel solution that eliminates the need for specific object models or depth\ninformation. Our method utilises score-based diffusion models to generate\nobject pose hypotheses to model the distribution of possible poses for the\nobject. Unlike previous methods that rely on costly trained likelihood\nestimators to remove outliers before pose aggregation using mean pooling, we\nintroduce a simpler approach using Mean Shift to estimate the mode of the\ndistribution as the final pose estimate. Our approach outperforms the current\nstate-of-the-art on the REAL275 dataset by a significant margin.\n", "date": "2024-12-16", "categories": []}
{"id": "2412.04279", "title": "Targeted Hard Sample Synthesis Based on Estimated Pose and Occlusion\n  Error for Improved Object Pose Estimation", "abstract": "  6D Object pose estimation is a fundamental component in robotics enabling\nefficient interaction with the environment. It is particularly challenging in\nbin-picking applications, where objects may be textureless and in difficult\nposes, and occlusion between objects of the same type may cause confusion even\nin well-trained models. We propose a novel method of hard example synthesis\nthat is model-agnostic, using existing simulators and the modeling of pose\nerror in both the camera-to-object viewsphere and occlusion space. Through\nevaluation of the model performance with respect to the distribution of object\nposes and occlusions, we discover regions of high error and generate realistic\ntraining samples to specifically target these regions. With our training\napproach, we demonstrate an improvement in correct detection rate of up to 20%\nacross several ROBI-dataset objects using state-of-the-art pose estimation\nmodels.\n", "date": "2024-12-05", "categories": []}
{"id": "2409.11512", "title": "Good Grasps Only: A data engine for self-supervised fine-tuning of pose\n  estimation using grasp poses for verification", "abstract": "  In this paper, we present a novel method for self-supervised fine-tuning of\npose estimation. Leveraging zero-shot pose estimation, our approach enables the\nrobot to automatically obtain training data without manual labeling. After pose\nestimation the object is grasped, and in-hand pose estimation is used for data\nvalidation. Our pipeline allows the system to fine-tune while the process is\nrunning, removing the need for a learning phase. The motivation behind our work\nlies in the need for rapid setup of pose estimation solutions. Specifically, we\naddress the challenging task of bin picking, which plays a pivotal role in\nflexible robotic setups. Our method is implemented on a robotics work-cell, and\ntested with four different objects. For all objects, our method increases the\nperformance and outperforms a state-of-the-art method trained on the CAD model\nof the objects. Project page available at gogoengine.github.io\n", "date": "2024-09-17", "categories": []}
{"id": "2410.01061", "title": "Pose Estimation of Buried Deep-Sea Objects using 3D Vision Deep Learning\n  Models", "abstract": "  We present an approach for pose and burial fraction estimation of debris\nfield barrels found on the seabed in the Southern California San Pedro Basin.\nOur computational workflow leverages recent advances in foundation models for\nsegmentation and a vision transformer-based approach to estimate the point\ncloud which defines the geometry of the barrel. We propose BarrelNet for\nestimating the 6-DOF pose and radius of buried barrels from the barrel point\nclouds as input. We train BarrelNet using synthetically generated barrel point\nclouds, and qualitatively demonstrate the potential of our approach using\nremotely operated vehicle (ROV) video footage of barrels found at a historic\ndump site. We compare our method to a traditional least squares fitting\napproach and show significant improvement according to our defined benchmarks.\n", "date": "2024-10-01", "categories": []}
{"id": "2412.02267", "title": "GSGTrack: Gaussian Splatting-Guided Object Pose Tracking from RGB Videos", "abstract": "  Tracking the 6DoF pose of unknown objects in monocular RGB video sequences is\ncrucial for robotic manipulation. However, existing approaches typically rely\non accurate depth information, which is non-trivial to obtain in real-world\nscenarios. Although depth estimation algorithms can be employed, geometric\ninaccuracy can lead to failures in RGBD-based pose tracking methods. To address\nthis challenge, we introduce GSGTrack, a novel RGB-based pose tracking\nframework that jointly optimizes geometry and pose. Specifically, we adopt 3D\nGaussian Splatting to create an optimizable 3D representation, which is learned\nsimultaneously with a graph-based geometry optimization to capture the object's\nappearance features and refine its geometry. However, the joint optimization\nprocess is susceptible to perturbations from noisy pose and geometry data.\nThus, we propose an object silhouette loss to address the issue of pixel-wise\nloss being overly sensitive to pose noise during tracking. To mitigate the\ngeometric ambiguities caused by inaccurate depth information, we propose a\ngeometry-consistent image pair selection strategy, which filters out\nlow-confidence pairs and ensures robust geometric optimization. Extensive\nexperiments on the OnePose and HO3D datasets demonstrate the effectiveness of\nGSGTrack in both 6DoF pose tracking and object reconstruction.\n", "date": "2024-12-03", "categories": []}
{"id": "2412.00835", "title": "Particle-based 6D Object Pose Estimation from Point Clouds using\n  Diffusion Models", "abstract": "  Object pose estimation from a single view remains a challenging problem. In\nparticular, partial observability, occlusions, and object symmetries eventually\nresult in pose ambiguity. To account for this multimodality, this work proposes\ntraining a diffusion-based generative model for 6D object pose estimation.\nDuring inference, the trained generative model allows for sampling multiple\nparticles, i.e., pose hypotheses. To distill this information into a single\npose estimate, we propose two novel and effective pose selection strategies\nthat do not require any additional training or computationally intensive\noperations. Moreover, while many existing methods for pose estimation primarily\nfocus on the image domain and only incorporate depth information for final pose\nrefinement, our model solely operates on point cloud data. The model thereby\nleverages recent advancements in point cloud processing and operates upon an\nSE(3)-equivariant latent space that forms the basis for the particle selection\nstrategies and allows for improved inference times. Our thorough experimental\nresults demonstrate the competitive performance of our approach on the Linemod\ndataset and showcase the effectiveness of our design choices. Code is available\nat https://github.com/zitronian/6DPoseDiffusion .\n", "date": "2024-12-01", "categories": []}
{"id": "2412.01052", "title": "CRISP: Object Pose and Shape Estimation with Test-Time Adaptation", "abstract": "  We consider the problem of estimating object pose and shape from an RGB-D\nimage. Our first contribution is to introduce CRISP, a category-agnostic object\npose and shape estimation pipeline. The pipeline implements an encoder-decoder\nmodel for shape estimation. It uses FiLM-conditioning for implicit shape\nreconstruction and a DPT-based network for estimating pose-normalized points\nfor pose estimation. As a second contribution, we propose an optimization-based\npose and shape corrector that can correct estimation errors caused by a domain\ngap. Observing that the shape decoder is well behaved in the convex hull of\nknown shapes, we approximate the shape decoder with an active shape model, and\nshow that this reduces the shape correction problem to a constrained linear\nleast squares problem, which can be solved efficiently by an interior point\nalgorithm. Third, we introduce a self-training pipeline to perform\nself-supervised domain adaptation of CRISP. The self-training is based on a\ncorrect-and-certify approach, which leverages the corrector to generate\npseudo-labels at test time, and uses them to self-train CRISP. We demonstrate\nCRISP (and the self-training) on YCBV, SPE3R, and NOCS datasets. CRISP shows\nhigh performance on all the datasets. Moreover, our self-training is capable of\nbridging a large domain gap. Finally, CRISP also shows an ability to generalize\nto unseen objects. Code and pre-trained models will be available on\nhttps://web.mit.edu/sparklab/research/crisp_object_pose_shape/.\n", "date": "2024-12-02", "categories": []}
{"id": "2411.17174", "title": "GMFlow: Global Motion-Guided Recurrent Flow for 6D Object Pose\n  Estimation", "abstract": "  6D object pose estimation is crucial for robotic perception and precise\nmanipulation. Occlusion and incomplete object visibility are common challenges\nin this task, but existing pose refinement methods often struggle to handle\nthese issues effectively. To tackle this problem, we propose a global\nmotion-guided recurrent flow estimation method called GMFlow for pose\nestimation. GMFlow overcomes local ambiguities caused by occlusion or missing\nparts by seeking global explanations. We leverage the object's structural\ninformation to extend the motion of visible parts of the rigid body to its\ninvisible regions. Specifically, we capture global contextual information\nthrough a linear attention mechanism and guide local motion information to\ngenerate global motion estimates. Furthermore, we introduce object shape\nconstraints in the flow iteration process, making flow estimation suitable for\npose estimation scenarios. Experiments on the LM-O and YCB-V datasets\ndemonstrate that our method outperforms existing techniques in accuracy while\nmaintaining competitive computational efficiency.\n", "date": "2024-11-26", "categories": []}
{"id": "2411.15860", "title": "Generalizable Single-view Object Pose Estimation by Two-side Generating\n  and Matching", "abstract": "  In this paper, we present a novel generalizable object pose estimation method\nto determine the object pose using only one RGB image. Unlike traditional\napproaches that rely on instance-level object pose estimation and necessitate\nextensive training data, our method offers generalization to unseen objects\nwithout extensive training, operates with a single reference image of the\nobject, and eliminates the need for 3D object models or multiple views of the\nobject. These characteristics are achieved by utilizing a diffusion model to\ngenerate novel-view images and conducting a two-sided matching on these\ngenerated images. Quantitative experiments demonstrate the superiority of our\nmethod over existing pose estimation techniques across both synthetic and\nreal-world datasets. Remarkably, our approach maintains strong performance even\nin scenarios with significant viewpoint changes, highlighting its robustness\nand versatility in challenging conditions. The code will be re leased at\nhttps://github.com/scy639/Gen2SM.\n", "date": "2024-11-24", "categories": []}
{"id": "2411.16668", "title": "Diffusion Features for Zero-Shot 6DoF Object Pose Estimation", "abstract": "  Zero-shot object pose estimation enables the retrieval of object poses from\nimages without necessitating object-specific training. In recent approaches\nthis is facilitated by vision foundation models (VFM), which are pre-trained\nmodels that are effectively general-purpose feature extractors. The\ncharacteristics exhibited by these VFMs vary depending on the training data,\nnetwork architecture, and training paradigm. The prevailing choice in this\nfield are self-supervised Vision Transformers (ViT). This study assesses the\ninfluence of Latent Diffusion Model (LDM) backbones on zero-shot pose\nestimation. In order to facilitate a comparison between the two families of\nmodels on a common ground we adopt and modify a recent approach. Therefore, a\ntemplate-based multi-staged method for estimating poses in a zero-shot fashion\nusing LDMs is presented. The efficacy of the proposed approach is empirically\nevaluated on three standard datasets for object-specific 6DoF pose estimation.\nThe experiments demonstrate an Average Recall improvement of up to 27% over the\nViT baseline. The source code is available at: https://github.com/BvG1993/DZOP.\n", "date": "2024-11-25", "categories": []}
{"id": "2411.14002", "title": "SEMPose: A Single End-to-end Network for Multi-object Pose Estimation", "abstract": "  In computer vision, estimating the six-degree-of-freedom pose from an RGB\nimage is a fundamental task. However, this task becomes highly challenging in\nmulti-object scenes. Currently, the best methods typically employ an indirect\nstrategy, which identifies 2D and 3D correspondences, and then solves with the\nPerspective-n-Points method. Yet, this approach cannot be trained end-to-end.\nDirect methods, on the other hand, suffer from lower accuracy due to challenges\nsuch as varying object sizes and occlusions. To address these issues, we\npropose SEMPose, an end-to-end multi-object pose estimation network. SEMPose\nutilizes a well-designed texture-shape guided feature pyramid network,\neffectively tackling the challenge of object size variations. Additionally, it\nemploys an iterative refinement head structure, progressively regressing\nrotation and translation separately to enhance estimation accuracy. During\ntraining, we alleviate the impact of occlusion by selecting positive samples\nfrom visible parts. Experimental results demonstrate that SEMPose can perform\ninference at 32 FPS without requiring inputs other than the RGB image. It can\naccurately estimate the poses of multiple objects in real time, with inference\ntime unaffected by the number of target objects. On the LM-O and YCB-V\ndatasets, our method outperforms other RGB-based single-model methods,\nachieving higher accuracy. Even when compared with multi-model methods and\napproaches that use additional refinement, our results remain competitive.\n", "date": "2024-11-21", "categories": []}
{"id": "2411.03807", "title": "GS2Pose: Two-stage 6D Object Pose Estimation Guided by Gaussian\n  Splatting", "abstract": "  This paper proposes a new method for accurate and robust 6D pose estimation\nof novel objects, named GS2Pose. By introducing 3D Gaussian splatting, GS2Pose\ncan utilize the reconstruction results without requiring a high-quality CAD\nmodel, which means it only requires segmented RGBD images as input.\nSpecifically, GS2Pose employs a two-stage structure consisting of coarse\nestimation followed by refined estimation. In the coarse stage, a lightweight\nU-Net network with a polarization attention mechanism, called Pose-Net, is\ndesigned. By using the 3DGS model for supervised training, Pose-Net can\ngenerate NOCS images to compute a coarse pose. In the refinement stage, GS2Pose\nformulates a pose regression algorithm following the idea of reprojection or\nBundle Adjustment (BA), referred to as GS-Refiner. By leveraging Lie algebra to\nextend 3DGS, GS-Refiner obtains a pose-differentiable rendering pipeline that\nrefines the coarse pose by comparing the input images with the rendered images.\nGS-Refiner also selectively updates parameters in the 3DGS model to achieve\nenvironmental adaptation, thereby enhancing the algorithm's robustness and\nflexibility to illuminative variation, occlusion, and other challenging\ndisruptive factors. GS2Pose was evaluated through experiments conducted on the\nLineMod dataset, where it was compared with similar algorithms, yielding highly\ncompetitive results. The code for GS2Pose will soon be released on GitHub.\n", "date": "2024-11-06", "categories": []}
{"id": "2312.02985", "title": "FocalPose++: Focal Length and Object Pose Estimation via Render and\n  Compare", "abstract": "  We introduce FocalPose++, a neural render-and-compare method for jointly\nestimating the camera-object 6D pose and camera focal length given a single RGB\ninput image depicting a known object. The contributions of this work are\nthreefold. First, we derive a focal length update rule that extends an existing\nstate-of-the-art render-and-compare 6D pose estimator to address the joint\nestimation task. Second, we investigate several different loss functions for\njointly estimating the object pose and focal length. We find that a combination\nof direct focal length regression with a reprojection loss disentangling the\ncontribution of translation, rotation, and focal length leads to improved\nresults. Third, we explore the effect of different synthetic training data on\nthe performance of our method. Specifically, we investigate different\ndistributions used for sampling object's 6D pose and camera's focal length when\nrendering the synthetic images, and show that parametric distribution fitted on\nreal training data works the best. We show results on three challenging\nbenchmark datasets that depict known 3D models in uncontrolled settings. We\ndemonstrate that our focal length and 6D pose estimates have lower error than\nthe existing state-of-the-art methods.\n", "date": "2023-11-15", "categories": []}
{"id": "2411.00543", "title": "3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction", "abstract": "  Determining the 3D orientations of an object in an image, known as\nsingle-image pose estimation, is a crucial task in 3D vision applications.\nExisting methods typically learn 3D rotations parametrized in the spatial\ndomain using Euler angles or quaternions, but these representations often\nintroduce discontinuities and singularities. SO(3)-equivariant networks enable\nthe structured capture of pose patterns with data-efficient learning, but the\nparametrizations in spatial domain are incompatible with their architecture,\nparticularly spherical CNNs, which operate in the frequency domain to enhance\ncomputational efficiency. To overcome these issues, we propose a\nfrequency-domain approach that directly predicts Wigner-D coefficients for 3D\nrotation regression, aligning with the operations of spherical CNNs. Our\nSO(3)-equivariant pose harmonics predictor overcomes the limitations of spatial\nparameterizations, ensuring consistent pose estimation under arbitrary\nrotations. Trained with a frequency-domain regression loss, our method achieves\nstate-of-the-art results on benchmarks such as ModelNet10-SO(3) and PASCAL3D+,\nwith significant improvements in accuracy, robustness, and data efficiency.\n", "date": "2024-11-01", "categories": []}
{"id": "2410.07801", "title": "LucidGrasp: Robotic Framework for Autonomous Manipulation of Laboratory\n  Equipment with Different Degrees of Transparency via 6D Pose Estimation", "abstract": "  Many modern robotic systems operate autonomously, however they often lack the\nability to accurately analyze the environment and adapt to changing external\nconditions, while teleoperation systems often require special operator skills.\nIn the field of laboratory automation, the number of automated processes is\ngrowing, however such systems are usually developed to perform specific tasks.\nIn addition, many of the objects used in this field are transparent, making it\ndifficult to analyze them using visual channels. The contributions of this work\ninclude the development of a robotic framework with autonomous mode for\nmanipulating liquid-filled objects with different degrees of transparency in\ncomplex pose combinations. The conducted experiments demonstrated the\nrobustness of the designed visual perception system to accurately estimate\nobject poses for autonomous manipulation, and confirmed the performance of the\nalgorithms in dexterous operations such as liquid dispensing. The proposed\nrobotic framework can be applied for laboratory automation, since it allows\nsolving the problem of performing non-trivial manipulation tasks with the\nanalysis of object poses of varying degrees of transparency and liquid levels,\nrequiring high accuracy and repeatability.\n", "date": "2024-10-10", "categories": []}
{"id": "2305.18947", "title": "A Probabilistic Rotation Representation for Symmetric Shapes With an\n  Efficiently Computable Bingham Loss Function", "abstract": "  In recent years, a deep learning framework has been widely used for object\npose estimation. While quaternion is a common choice for rotation\nrepresentation, it cannot represent the ambiguity of the observation. In order\nto handle the ambiguity, the Bingham distribution is one promising solution.\nHowever, it requires complicated calculation when yielding the negative\nlog-likelihood (NLL) loss. An alternative easy-to-implement loss function has\nbeen proposed to avoid complex computations but has difficulty expressing\nsymmetric distribution. In this paper, we introduce a fast-computable and\neasy-to-implement NLL loss function for Bingham distribution. We also create\nthe inference network and show that our loss function can capture the symmetric\nproperty of target objects from their point clouds.\n", "date": "2023-05-30", "categories": []}
{"id": "2310.13320", "title": "CylinderTag: An Accurate and Flexible Marker for Cylinder-Shape Objects\n  Pose Estimation Based on Projective Invariants", "abstract": "  High-precision pose estimation based on visual markers has been a thriving\nresearch topic in the field of computer vision. However, the suitability of\ntraditional flat markers on curved objects is limited due to the diverse shapes\nof curved surfaces, which hinders the development of high-precision pose\nestimation for curved objects. Therefore, this paper proposes a novel visual\nmarker called CylinderTag, which is designed for developable curved surfaces\nsuch as cylindrical surfaces. CylinderTag is a cyclic marker that can be firmly\nattached to objects with a cylindrical shape. Leveraging the manifold\nassumption, the cross-ratio in projective invariance is utilized for encoding\nin the direction of zero curvature on the surface. Additionally, to facilitate\nthe usage of CylinderTag, we propose a heuristic search-based marker generator\nand a high-performance recognizer as well. Moreover, an all-encompassing\nevaluation of CylinderTag properties is conducted by means of extensive\nexperimentation, covering detection rate, detection speed, dictionary size,\nlocalization jitter, and pose estimation accuracy. CylinderTag showcases\nsuperior detection performance from varying view angles in comparison to\ntraditional visual markers, accompanied by higher localization accuracy.\nFurthermore, CylinderTag boasts real-time detection capability and an extensive\nmarker dictionary, offering enhanced versatility and practicality in a wide\nrange of applications. Experimental results demonstrate that the CylinderTag is\na highly promising visual marker for use on cylindrical-like surfaces, thus\noffering important guidance for future research on high-precision visual\nlocalization of cylinder-shaped objects. The code is available at:\nhttps://github.com/wsakobe/CylinderTag.\n", "date": "2023-10-20", "categories": []}
{"id": "2402.12647", "title": "DiffusionNOCS: Managing Symmetry and Uncertainty in Sim2Real Multi-Modal\n  Category-level Pose Estimation", "abstract": "  This paper addresses the challenging problem of category-level pose\nestimation. Current state-of-the-art methods for this task face challenges when\ndealing with symmetric objects and when attempting to generalize to new\nenvironments solely through synthetic data training. In this work, we address\nthese challenges by proposing a probabilistic model that relies on diffusion to\nestimate dense canonical maps crucial for recovering partial object shapes as\nwell as establishing correspondences essential for pose estimation.\nFurthermore, we introduce critical components to enhance performance by\nleveraging the strength of the diffusion models with multi-modal input\nrepresentations. We demonstrate the effectiveness of our method by testing it\non a range of real datasets. Despite being trained solely on our generated\nsynthetic data, our approach achieves state-of-the-art performance and\nunprecedented generalization qualities, outperforming baselines, even those\nspecifically trained on the target domain.\n", "date": "2024-02-20", "categories": []}
{"id": "2305.17828", "title": "Counter-Hypothetical Particle Filters for Single Object Pose Tracking", "abstract": "  Particle filtering is a common technique for six degrees of freedom (6D) pose\nestimation due to its ability to tractably represent belief over object pose.\nHowever, the particle filter is prone to particle deprivation due to the\nhigh-dimensional nature of 6D pose. When particle deprivation occurs, it can\ncause mode collapse of the underlying belief distribution during importance\nsampling. If the region surrounding the true state suffers from mode collapse,\nrecovering its belief is challenging since the area is no longer represented in\nthe probability mass formed by the particles. Previous methods mitigate this\nproblem by randomizing and resetting particles in the belief distribution, but\ndetermining the frequency of reinvigoration has relied on hand-tuning abstract\nheuristics. In this paper, we estimate the necessary reinvigoration rate at\neach time step by introducing a Counter-Hypothetical likelihood function, which\nis used alongside the standard likelihood. Inspired by the notions of\nplausibility and implausibility from Evidential Reasoning, the addition of our\nCounter-Hypothetical likelihood function assigns a level of doubt to each\nparticle. The competing cumulative values of confidence and doubt across the\nparticle set are used to estimate the level of failure within the filter, in\norder to determine the portion of particles to be reinvigorated. We demonstrate\nthe effectiveness of our method on the rigid body object 6D pose tracking task.\n", "date": "2023-05-28", "categories": []}
{"id": "2409.06240", "title": "Test-Time Certifiable Self-Supervision to Bridge the Sim2Real Gap in\n  Event-Based Satellite Pose Estimation", "abstract": "  Deep learning plays a critical role in vision-based satellite pose\nestimation. However, the scarcity of real data from the space environment means\nthat deep models need to be trained using synthetic data, which raises the\nSim2Real domain gap problem. A major cause of the Sim2Real gap are novel\nlighting conditions encountered during test time. Event sensors have been shown\nto provide some robustness against lighting variations in vision-based pose\nestimation. However, challenging lighting conditions due to strong directional\nlight can still cause undesirable effects in the output of commercial\noff-the-shelf event sensors, such as noisy/spurious events and inhomogeneous\nevent densities on the object. Such effects are non-trivial to simulate in\nsoftware, thus leading to Sim2Real gap in the event domain. To close the\nSim2Real gap in event-based satellite pose estimation, the paper proposes a\ntest-time self-supervision scheme with a certifier module. Self-supervision is\nenabled by an optimisation routine that aligns a dense point cloud of the\npredicted satellite pose with the event data to attempt to rectify the\ninaccurately estimated pose. The certifier attempts to verify the corrected\npose, and only certified test-time inputs are backpropagated via implicit\ndifferentiation to refine the predicted landmarks, thus improving the pose\nestimates and closing the Sim2Real gap. Results show that the our method\noutperforms established test-time adaptation schemes.\n", "date": "2024-09-10", "categories": []}
{"id": "2409.19986", "title": "SuperPose: Improved 6D Pose Estimation with Robust Tracking and\n  Mask-Free Initialization", "abstract": "  We developed a robust solution for real-time 6D object detection in\nindustrial applications by integrating FoundationPose, SAM2, and LightGlue,\neliminating the need for retraining. Our approach addresses two key challenges:\nthe requirement for an initial object mask in the first frame in FoundationPose\nand issues with tracking loss and automatic rotation for symmetric objects. The\nalgorithm requires only a CAD model of the target object, with the user\nclicking on its location in the live feed during the initial setup. Once set,\nthe algorithm automatically saves a reference image of the object and, in\nsubsequent runs, employs LightGlue for feature matching between the object and\nthe real-time scene, providing an initial prompt for detection. Tested on the\nYCB dataset and industrial components such as bleach cleanser and gears, the\nalgorithm demonstrated reliable 6D detection and tracking. By integrating SAM2\nand FoundationPose, we effectively mitigated common limitations such as the\nproblem of tracking loss, ensuring continuous and accurate tracking under\nchallenging conditions like occlusion or rapid movement.\n", "date": "2024-09-30", "categories": []}
{"id": "2410.12679", "title": "Optimizing Multi-Task Learning for Accurate Spacecraft Pose Estimation", "abstract": "  Accurate satellite pose estimation is crucial for autonomous guidance,\nnavigation, and control (GNC) systems in in-orbit servicing (IOS) missions.\nThis paper explores the impact of different tasks within a multi-task learning\n(MTL) framework for satellite pose estimation using monocular images. By\nintegrating tasks such as direct pose estimation, keypoint prediction, object\nlocalization, and segmentation into a single network, the study aims to\nevaluate the reciprocal influence between tasks by testing different multi-task\nconfigurations thanks to the modularity of the convolutional neural network\n(CNN) used in this work. The trends of mutual bias between the analyzed tasks\nare found by employing different weighting strategies to further test the\nrobustness of the findings. A synthetic dataset was developed to train and test\nthe MTL network. Results indicate that direct pose estimation and heatmap-based\npose estimation positively influence each other in general, while both the\nbounding box and segmentation tasks do not provide significant contributions\nand tend to degrade the overall estimation accuracy.\n", "date": "2024-10-16", "categories": []}
{"id": "2403.18791", "title": "Object Pose Estimation via the Aggregation of Diffusion Features", "abstract": "  Estimating the pose of objects from images is a crucial task of 3D scene\nunderstanding, and recent approaches have shown promising results on very large\nbenchmarks. However, these methods experience a significant performance drop\nwhen dealing with unseen objects. We believe that it results from the limited\ngeneralizability of image features. To address this problem, we have an\nin-depth analysis on the features of diffusion models, e.g. Stable Diffusion,\nwhich hold substantial potential for modeling unseen objects. Based on this\nanalysis, we then innovatively introduce these diffusion features for object\npose estimation. To achieve this, we propose three distinct architectures that\ncan effectively capture and aggregate diffusion features of different\ngranularity, greatly improving the generalizability of object pose estimation.\nOur approach outperforms the state-of-the-art methods by a considerable margin\non three popular benchmark datasets, LM, O-LM, and T-LESS. In particular, our\nmethod achieves higher accuracy than the previous best arts on unseen objects:\n97.9% vs. 93.5% on Unseen LM, 85.9% vs. 76.3% on Unseen O-LM, showing the\nstrong generalizability of our method. Our code is released at\nhttps://github.com/Tianfu18/diff-feats-pose.\n", "date": "2024-03-27", "categories": []}
{"id": "2410.13465", "title": "Object Pose Estimation Using Implicit Representation For Transparent\n  Objects", "abstract": "  Object pose estimation is a prominent task in computer vision. The object\npose gives the orientation and translation of the object in real-world space,\nwhich allows various applications such as manipulation, augmented reality, etc.\nVarious objects exhibit different properties with light, such as reflections,\nabsorption, etc. This makes it challenging to understand the object's structure\nin RGB and depth channels. Recent research has been moving toward\nlearning-based methods, which provide a more flexible and generalizable\napproach to object pose estimation utilizing deep learning. One such approach\nis the render-and-compare method, which renders the object from multiple views\nand compares it against the given 2D image, which often requires an object\nrepresentation in the form of a CAD model. We reason that the synthetic texture\nof the CAD model may not be ideal for rendering and comparing operations. We\nshowed that if the object is represented as an implicit (neural) representation\nin the form of Neural Radiance Field (NeRF), it exhibits a more realistic\nrendering of the actual scene and retains the crucial spatial features, which\nmakes the comparison more versatile. We evaluated our NeRF implementation of\nthe render-and-compare method on transparent datasets and found that it\nsurpassed the current state-of-the-art results.\n", "date": "2024-10-17", "categories": []}
{"id": "2410.09010", "title": "CVAM-Pose: Conditional Variational Autoencoder for Multi-Object\n  Monocular Pose Estimation", "abstract": "  Estimating rigid objects' poses is one of the fundamental problems in\ncomputer vision, with a range of applications across automation and augmented\nreality. Most existing approaches adopt one network per object class strategy,\ndepend heavily on objects' 3D models, depth data, and employ a time-consuming\niterative refinement, which could be impractical for some applications. This\npaper presents a novel approach, CVAM-Pose, for multi-object monocular pose\nestimation that addresses these limitations. The CVAM-Pose method employs a\nlabel-embedded conditional variational autoencoder network, to implicitly\nabstract regularised representations of multiple objects in a single\nlow-dimensional latent space. This autoencoding process uses only images\ncaptured by a projective camera and is robust to objects' occlusion and scene\nclutter. The classes of objects are one-hot encoded and embedded throughout the\nnetwork. The proposed label-embedded pose regression strategy interprets the\nlearnt latent space representations utilising continuous pose representations.\nAblation tests and systematic evaluations demonstrate the scalability and\nefficiency of the CVAM-Pose method for multi-object scenarios. The proposed\nCVAM-Pose outperforms competing latent space approaches. For example, it is\nrespectively 25% and 20% better than AAE and Multi-Path methods, when evaluated\nusing the $\\mathrm{AR_{VSD}}$ metric on the Linemod-Occluded dataset. It also\nachieves results somewhat comparable to methods reliant on 3D models reported\nin BOP challenges. Code available: https://github.com/JZhao12/CVAM-Pose\n", "date": "2024-10-11", "categories": []}
{"id": "2410.04298", "title": "Test-Time Adaptation for Keypoint-Based Spacecraft Pose Estimation Based\n  on Predicted-View Synthesis", "abstract": "  Due to the difficulty of replicating the real conditions during training,\nsupervised algorithms for spacecraft pose estimation experience a drop in\nperformance when trained on synthetic data and applied to real operational\ndata. To address this issue, we propose a test-time adaptation approach that\nleverages the temporal redundancy between images acquired during close\nproximity operations. Our approach involves extracting features from sequential\nspacecraft images, estimating their poses, and then using this information to\nsynthesise a reconstructed view. We establish a self-supervised learning\nobjective by comparing the synthesised view with the actual one. During\ntraining, we supervise both pose estimation and image synthesis, while at\ntest-time, we optimise the self-supervised objective. Additionally, we\nintroduce a regularisation loss to prevent solutions that are not consistent\nwith the keypoint structure of the spacecraft. Our code is available at:\nhttps://github.com/JotaBravo/spacecraft-tta.\n", "date": "2024-10-05", "categories": []}
{"id": "2305.17934", "title": "ZeroPose: CAD-Prompted Zero-shot Object 6D Pose Estimation in Cluttered\n  Scenes", "abstract": "  Many robotics and industry applications have a high demand for the capability\nto estimate the 6D pose of novel objects from the cluttered scene. However,\nexisting classic pose estimation methods are object-specific, which can only\nhandle the specific objects seen during training. When applied to a novel\nobject, these methods necessitate a cumbersome onboarding process, which\ninvolves extensive dataset preparation and model retraining. The extensive\nduration and resource consumption of onboarding limit their practicality in\nreal-world applications. In this paper, we introduce ZeroPose, a novel\nzero-shot framework that performs pose estimation following a\nDiscovery-Orientation-Registration (DOR) inference pipeline. This framework\ngeneralizes to novel objects without requiring model retraining. Given the CAD\nmodel of a novel object, ZeroPose enables in seconds onboarding time to extract\nvisual and geometric embeddings from the CAD model as a prompt. With the\nprompting of the above embeddings, DOR can discover all related instances and\nestimate their 6D poses without additional human interaction or presupposing\nscene conditions. Compared with existing zero-shot methods solved by the\nrender-and-compare paradigm, the DOR pipeline formulates the object pose\nestimation into a feature-matching problem, which avoids time-consuming online\nrendering and improves efficiency. Experimental results on the seven datasets\nshow that ZeroPose as a zero-shot method achieves comparable performance with\nobject-specific training methods and outperforms the state-of-the-art zero-shot\nmethod with 50x inference speed improvement.\n", "date": "2023-05-29", "categories": []}
{"id": "2409.16600", "title": "FAFA: Frequency-Aware Flow-Aided Self-Supervision for Underwater Object\n  Pose Estimation", "abstract": "  Although methods for estimating the pose of objects in indoor scenes have\nachieved great success, the pose estimation of underwater objects remains\nchallenging due to difficulties brought by the complex underwater environment,\nsuch as degraded illumination, blurring, and the substantial cost of obtaining\nreal annotations. In response, we introduce FAFA, a Frequency-Aware Flow-Aided\nself-supervised framework for 6D pose estimation of unmanned underwater\nvehicles (UUVs). Essentially, we first train a frequency-aware flow-based pose\nestimator on synthetic data, where an FFT-based augmentation approach is\nproposed to facilitate the network in capturing domain-invariant features and\ntarget domain styles from a frequency perspective. Further, we perform\nself-supervised training by enforcing flow-aided multi-level consistencies to\nadapt it to the real-world underwater environment. Our framework relies solely\non the 3D model and RGB images, alleviating the need for any real pose\nannotations or other-modality data like depths. We evaluate the effectiveness\nof FAFA on common underwater object pose benchmarks and showcase significant\nperformance improvements compared to state-of-the-art methods. Code is\navailable at github.com/tjy0703/FAFA.\n", "date": "2024-09-25", "categories": []}
{"id": "2409.15727", "title": "LaPose: Laplacian Mixture Shape Modeling for RGB-Based Category-Level\n  Object Pose Estimation", "abstract": "  While RGBD-based methods for category-level object pose estimation hold\npromise, their reliance on depth data limits their applicability in diverse\nscenarios. In response, recent efforts have turned to RGB-based methods;\nhowever, they face significant challenges stemming from the absence of depth\ninformation. On one hand, the lack of depth exacerbates the difficulty in\nhandling intra-class shape variation, resulting in increased uncertainty in\nshape predictions. On the other hand, RGB-only inputs introduce inherent scale\nambiguity, rendering the estimation of object size and translation an ill-posed\nproblem. To tackle these challenges, we propose LaPose, a novel framework that\nmodels the object shape as the Laplacian mixture model for Pose estimation. By\nrepresenting each point as a probabilistic distribution, we explicitly quantify\nthe shape uncertainty. LaPose leverages both a generalized 3D information\nstream and a specialized feature stream to independently predict the Laplacian\ndistribution for each point, capturing different aspects of object geometry.\nThese two distributions are then integrated as a Laplacian mixture model to\nestablish the 2D-3D correspondences, which are utilized to solve the pose via\nthe PnP module. In order to mitigate scale ambiguity, we introduce a\nscale-agnostic representation for object size and translation, enhancing\ntraining efficiency and overall robustness. Extensive experiments on the NOCS\ndatasets validate the effectiveness of LaPose, yielding state-of-the-art\nperformance in RGB-based category-level object pose estimation. Codes are\nreleased at https://github.com/lolrudy/LaPose\n", "date": "2024-09-24", "categories": []}
{"id": "2409.12720", "title": "FAST GDRNPP: Improving the Speed of State-of-the-Art 6D Object Pose\n  Estimation", "abstract": "  6D object pose estimation involves determining the three-dimensional\ntranslation and rotation of an object within a scene and relative to a chosen\ncoordinate system. This problem is of particular interest for many practical\napplications in industrial tasks such as quality control, bin picking, and\nrobotic manipulation, where both speed and accuracy are critical for real-world\ndeployment. Current models, both classical and deep-learning-based, often\nstruggle with the trade-off between accuracy and latency. Our research focuses\non enhancing the speed of a prominent state-of-the-art deep learning model,\nGDRNPP, while keeping its high accuracy. We employ several techniques to reduce\nthe model size and improve inference time. These techniques include using\nsmaller and quicker backbones, pruning unnecessary parameters, and distillation\nto transfer knowledge from a large, high-performing model to a smaller, more\nefficient student model. Our findings demonstrate that the proposed\nconfiguration maintains accuracy comparable to the state-of-the-art while\nsignificantly improving inference time. This advancement could lead to more\nefficient and practical applications in various industrial scenarios, thereby\nenhancing the overall applicability of 6D Object Pose Estimation models in\nreal-world settings.\n", "date": "2024-09-18", "categories": []}
{"id": "2409.11819", "title": "End-to-End Probabilistic Geometry-Guided Regression for 6DoF Object Pose\n  Estimation", "abstract": "  6D object pose estimation is the problem of identifying the position and\norientation of an object relative to a chosen coordinate system, which is a\ncore technology for modern XR applications. State-of-the-art 6D object pose\nestimators directly predict an object pose given an object observation. Due to\nthe ill-posed nature of the pose estimation problem, where multiple different\nposes can correspond to a single observation, generating additional plausible\nestimates per observation can be valuable. To address this, we reformulate the\nstate-of-the-art algorithm GDRNPP and introduce EPRO-GDR (End-to-End\nProbabilistic Geometry-Guided Regression). Instead of predicting a single pose\nper detection, we estimate a probability density distribution of the pose.\nUsing the evaluation procedure defined by the BOP (Benchmark for 6D Object Pose\nEstimation) Challenge, we test our approach on four of its core datasets and\ndemonstrate superior quantitative results for EPRO-GDR on LM-O, YCB-V, and\nITODD. Our probabilistic solution shows that predicting a pose distribution\ninstead of a single pose can improve state-of-the-art single-view pose\nestimation while providing the additional benefit of being able to sample\nmultiple meaningful pose candidates.\n", "date": "2024-09-18", "categories": []}
{"id": "2210.03437", "title": "PCKRF: Point Cloud Completion and Keypoint Refinement With Fusion Data\n  for 6D Pose Estimation", "abstract": "  Some robust point cloud registration approaches with controllable pose\nrefinement magnitude, such as ICP and its variants, are commonly used to\nimprove 6D pose estimation accuracy. However, the effectiveness of these\nmethods gradually diminishes with the advancement of deep learning techniques\nand the enhancement of initial pose accuracy, primarily due to their lack of\nspecific design for pose refinement. In this paper, we propose Point Cloud\nCompletion and Keypoint Refinement with Fusion Data (PCKRF), a new pose\nrefinement pipeline for 6D pose estimation. The pipeline consists of two steps.\nFirst, it completes the input point clouds via a novel pose-sensitive point\ncompletion network. The network uses both local and global features with pose\ninformation during point completion. Then, it registers the completed object\npoint cloud with the corresponding target point cloud by our proposed Color\nsupported Iterative KeyPoint (CIKP) method. The CIKP method introduces color\ninformation into registration and registers a point cloud around each keypoint\nto increase stability. The PCKRF pipeline can be integrated with existing\npopular 6D pose estimation methods, such as the full flow bidirectional fusion\nnetwork, to further improve their pose estimation accuracy. Experiments\ndemonstrate that our method exhibits superior stability compared to existing\napproaches when optimizing initial poses with relatively high precision.\nNotably, the results indicate that our method effectively complements most\nexisting pose estimation techniques, leading to improved performance in most\ncases. Furthermore, our method achieves promising results even in challenging\nscenarios involving textureless and symmetrical objects. Our source code is\navailable at https://github.com/zhanhz/KRF.\n", "date": "2022-10-07", "categories": []}
{"id": "2409.06683", "title": "Alignist: CAD-Informed Orientation Distribution Estimation by Fusing\n  Shape and Correspondences", "abstract": "  Object pose distribution estimation is crucial in robotics for better path\nplanning and handling of symmetric objects. Recent distribution estimation\napproaches employ contrastive learning-based approaches by maximizing the\nlikelihood of a single pose estimate in the absence of a CAD model. We propose\na pose distribution estimation method leveraging symmetry respecting\ncorrespondence distributions and shape information obtained using a CAD model.\nContrastive learning-based approaches require an exhaustive amount of training\nimages from different viewpoints to learn the distribution properly, which is\nnot possible in realistic scenarios. Instead, we propose a pipeline that can\nleverage correspondence distributions and shape information from the CAD model,\nwhich are later used to learn pose distributions. Besides, having access to\npose distribution based on correspondences before learning pose distributions\nconditioned on images, can help formulate the loss between distributions. The\nprior knowledge of distribution also helps the network to focus on getting\nsharper modes instead. With the CAD prior, our approach converges much faster\nand learns distribution better by focusing on learning sharper distribution\nnear all the valid modes, unlike contrastive approaches, which focus on a\nsingle mode at a time. We achieve benchmark results on SYMSOL-I and T-Less\ndatasets.\n", "date": "2024-09-10", "categories": []}
{"id": "2402.05610", "title": "Extending 6D Object Pose Estimators for Stereo Vision", "abstract": "  Estimating the 6D pose of objects accurately, quickly, and robustly remains a\ndifficult task. However, recent methods for directly regressing poses from RGB\nimages using dense features have achieved state-of-the-art results. Stereo\nvision, which provides an additional perspective on the object, can help reduce\npose ambiguity and occlusion. Moreover, stereo can directly infer the distance\nof an object, while mono-vision requires internalized knowledge of the object's\nsize. To extend the state-of-the-art in 6D object pose estimation to stereo, we\ncreated a BOP compatible stereo version of the YCB-V dataset. Our method\noutperforms state-of-the-art 6D pose estimation algorithms by utilizing stereo\nvision and can easily be adopted for other dense feature-based algorithms.\n", "date": "2024-02-08", "categories": []}
{"id": "2409.05413", "title": "From Words to Poses: Enhancing Novel Object Pose Estimation with Vision\n  Language Models", "abstract": "  Robots are increasingly envisioned to interact in real-world scenarios, where\nthey must continuously adapt to new situations. To detect and grasp novel\nobjects, zero-shot pose estimators determine poses without prior knowledge.\nRecently, vision language models (VLMs) have shown considerable advances in\nrobotics applications by establishing an understanding between language input\nand image input. In our work, we take advantage of VLMs zero-shot capabilities\nand translate this ability to 6D object pose estimation. We propose a novel\nframework for promptable zero-shot 6D object pose estimation using language\nembeddings. The idea is to derive a coarse location of an object based on the\nrelevancy map of a language-embedded NeRF reconstruction and to compute the\npose estimate with a point cloud registration method. Additionally, we provide\nan analysis of LERF's suitability for open-set object pose estimation. We\nexamine hyperparameters, such as activation thresholds for relevancy maps and\ninvestigate the zero-shot capabilities on an instance- and category-level.\nFurthermore, we plan to conduct robotic grasping experiments in a real-world\nsetting.\n", "date": "2024-09-09", "categories": []}
{"id": "2409.03556", "title": "MaskVal: Simple but Effective Uncertainty Quantification for 6D Pose\n  Estimation", "abstract": "  For the use of 6D pose estimation in robotic applications, reliable poses are\nof utmost importance to ensure a safe, reliable and predictable operational\nperformance. Despite these requirements, state-of-the-art 6D pose estimators\noften do not provide any uncertainty quantification for their pose estimates at\nall, or if they do, it has been shown that the uncertainty provided is only\nweakly correlated with the actual true error. To address this issue, we\ninvestigate a simple but effective uncertainty quantification, that we call\nMaskVal, which compares the pose estimates with their corresponding instance\nsegmentations by rendering and does not require any modification of the pose\nestimator itself. Despite its simplicity, MaskVal significantly outperforms a\nstate-of-the-art ensemble method on both a dataset and a robotic setup. We show\nthat by using MaskVal, the performance of a state-of-the-art 6D pose estimator\nis significantly improved towards a safe and reliable operation. In addition,\nwe propose a new and specific approach to compare and evaluate uncertainty\nquantification methods for 6D pose estimation in the context of robotic\nmanipulation.\n", "date": "2024-09-05", "categories": []}
{"id": "2409.02581", "title": "Object Gaussian for Monocular 6D Pose Estimation from Sparse Views", "abstract": "  Monocular object pose estimation, as a pivotal task in computer vision and\nrobotics, heavily depends on accurate 2D-3D correspondences, which often demand\ncostly CAD models that may not be readily available. Object 3D reconstruction\nmethods offer an alternative, among which recent advancements in 3D Gaussian\nSplatting (3DGS) afford a compelling potential. Yet its performance still\nsuffers and tends to overfit with fewer input views. Embracing this challenge,\nwe introduce SGPose, a novel framework for sparse view object pose estimation\nusing Gaussian-based methods. Given as few as ten views, SGPose generates a\ngeometric-aware representation by starting with a random cuboid initialization,\neschewing reliance on Structure-from-Motion (SfM) pipeline-derived geometry as\nrequired by traditional 3DGS methods. SGPose removes the dependence on CAD\nmodels by regressing dense 2D-3D correspondences between images and the\nreconstructed model from sparse input and random initialization, while the\ngeometric-consistent depth supervision and online synthetic view warping are\nkey to the success. Experiments on typical benchmarks, especially on the\nOcclusion LM-O dataset, demonstrate that SGPose outperforms existing methods\neven under sparse view constraints, under-scoring its potential in real-world\napplications.\n", "date": "2024-09-04", "categories": []}
{"id": "2408.05867", "title": "SABER-6D: Shape Representation Based Implicit Object Pose Estimation", "abstract": "  In this paper, we propose a novel encoder-decoder architecture, named SABER,\nto learn the 6D pose of the object in the embedding space by learning shape\nrepresentation at a given pose. This model enables us to learn pose by\nperforming shape representation at a target pose from RGB image input. We\nperform shape representation as an auxiliary task which helps us in learning\nrotations space for an object based on 2D images. An image encoder predicts the\nrotation in the embedding space and the DeepSDF based decoder learns to\nrepresent the object's shape at the given pose. As our approach is shape based,\nthe pipeline is suitable for any type of object irrespective of the symmetry.\nMoreover, we need only a CAD model of the objects to train SABER. Our pipeline\nis synthetic data based and can also handle symmetric objects without symmetry\nlabels and, thus, no additional labeled training data is needed. The\nexperimental evaluation shows that our method achieves close to benchmark\nresults for both symmetric objects and asymmetric objects on Occlusion-LineMOD,\nand T-LESS datasets.\n", "date": "2024-08-11", "categories": []}
{"id": "2303.16102", "title": "KeyMatchNet: Zero-Shot Pose Estimation in 3D Point Clouds by Generalized\n  Keypoint Matching", "abstract": "  In this paper, we present KeyMatchNet, a novel network for zero-shot pose\nestimation in 3D point clouds. Our method uses only depth information, making\nit more applicable for many industrial use cases, as color information is\nseldom available. The network is composed of two parallel components for\ncomputing object and scene features. The features are then combined to create\nmatches used for pose estimation. The parallel structure allows for\npre-processing of the individual parts, which decreases the run-time. Using a\nzero-shot network allows for a very short set-up time, as it is not necessary\nto train models for new objects. However, as the network is not trained for the\nspecific object, zero-shot pose estimation methods generally have lower\naccuracy compared with conventional methods. To address this, we reduce the\ncomplexity of the task by including the scenario information during training.\nThis is typically not feasible as collecting real data for new tasks\ndrastically increases the cost. However, for zero-shot pose estimation,\ntraining for new objects is not necessary and the expensive data collection can\nthus be performed only once. Our method is trained on 1,500 objects and is only\ntested on unseen objects. We demonstrate that the trained network can not only\naccurately estimate poses for novel objects, but also demonstrate the ability\nof the network on objects outside of the trained class. Test results are also\nshown on real data. We believe that the presented method is valuable for many\nreal-world scenarios. Project page available at keymatchnet.github.io\n", "date": "2023-03-28", "categories": []}
{"id": "2403.10683", "title": "GS-Pose: Generalizable Segmentation-based 6D Object Pose Estimation with\n  3D Gaussian Splatting", "abstract": "  This paper introduces GS-Pose, a unified framework for localizing and\nestimating the 6D pose of novel objects. GS-Pose begins with a set of posed RGB\nimages of a previously unseen object and builds three distinct representations\nstored in a database. At inference, GS-Pose operates sequentially by locating\nthe object in the input image, estimating its initial 6D pose using a retrieval\napproach, and refining the pose with a render-and-compare method. The key\ninsight is the application of the appropriate object representation at each\nstage of the process. In particular, for the refinement step, we leverage 3D\nGaussian splatting, a novel differentiable rendering technique that offers high\nrendering speed and relatively low optimization time. Off-the-shelf toolchains\nand commodity hardware, such as mobile phones, can be used to capture new\nobjects to be added to the database. Extensive evaluations on the LINEMOD and\nOnePose-LowTexture datasets demonstrate excellent performance, establishing the\nnew state-of-the-art. Project page: https://dingdingcai.github.io/gs-pose.\n", "date": "2024-03-15", "categories": []}
{"id": "2403.16400", "title": "ASDF: Assembly State Detection Utilizing Late Fusion by Integrating 6D\n  Pose Estimation", "abstract": "  In medical and industrial domains, providing guidance for assembly processes\ncan be critical to ensure efficiency and safety. Errors in assembly can lead to\nsignificant consequences such as extended surgery times and prolonged\nmanufacturing or maintenance times in industry. Assembly scenarios can benefit\nfrom in-situ augmented reality visualization, i.e., augmentations in close\nproximity to the target object, to provide guidance, reduce assembly times, and\nminimize errors. In order to enable in-situ visualization, 6D pose estimation\ncan be leveraged to identify the correct location for an augmentation. Existing\n6D pose estimation techniques primarily focus on individual objects and static\ncaptures. However, assembly scenarios have various dynamics, including\nocclusion during assembly and dynamics in the appearance of assembly objects.\nExisting work focus either on object detection combined with state detection,\nor focus purely on the pose estimation. To address the challenges of 6D pose\nestimation in combination with assembly state detection, our approach ASDF\nbuilds upon the strengths of YOLOv8, a real-time capable object detection\nframework. We extend this framework, refine the object pose, and fuse pose\nknowledge with network-detected pose information. Utilizing our late fusion in\nour Pose2State module results in refined 6D pose estimation and assembly state\ndetection. By combining both pose and state information, our Pose2State module\npredicts the final assembly state with precision. The evaluation of our ASDF\ndataset shows that our Pose2State module leads to an improved assembly state\ndetection and that the improvement of the assembly state further leads to a\nmore robust 6D pose estimation. Moreover, on the GBOT dataset, we outperform\nthe pure deep learning-based network and even outperform the hybrid and pure\ntracking-based approaches.\n", "date": "2024-03-25", "categories": []}
{"id": "2408.03225", "title": "Line-based 6-DoF Object Pose Estimation and Tracking With an Event\n  Camera", "abstract": "  Pose estimation and tracking of objects is a fundamental application in 3D\nvision. Event cameras possess remarkable attributes such as high dynamic range,\nlow latency, and resilience against motion blur, which enables them to address\nchallenging high dynamic range scenes or high-speed motion. These features make\nevent cameras an ideal complement over standard cameras for object pose\nestimation. In this work, we propose a line-based robust pose estimation and\ntracking method for planar or non-planar objects using an event camera.\nFirstly, we extract object lines directly from events, then provide an initial\npose using a globally-optimal Branch-and-Bound approach, where 2D-3D line\ncorrespondences are not known in advance. Subsequently, we utilize event-line\nmatching to establish correspondences between 2D events and 3D models.\nFurthermore, object poses are refined and continuously tracked by minimizing\nevent-line distances. Events are assigned different weights based on these\ndistances, employing robust estimation algorithms. To evaluate the precision of\nthe proposed methods in object pose estimation and tracking, we have devised\nand established an event-based moving object dataset. Compared against\nstate-of-the-art methods, the robustness and accuracy of our methods have been\nvalidated both on synthetic experiments and the proposed dataset. The source\ncode is available at https://github.com/Zibin6/LOPET.\n", "date": "2024-08-06", "categories": []}
{"id": "2403.14279", "title": "Zero123-6D: Zero-shot Novel View Synthesis for RGB Category-level 6D\n  Pose Estimation", "abstract": "  Estimating the pose of objects through vision is essential to make robotic\nplatforms interact with the environment. Yet, it presents many challenges,\noften related to the lack of flexibility and generalizability of\nstate-of-the-art solutions. Diffusion models are a cutting-edge neural\narchitecture transforming 2D and 3D computer vision, outlining remarkable\nperformances in zero-shot novel-view synthesis. Such a use case is particularly\nintriguing for reconstructing 3D objects. However, localizing objects in\nunstructured environments is rather unexplored. To this end, this work presents\nZero123-6D, the first work to demonstrate the utility of Diffusion Model-based\nnovel-view-synthesizers in enhancing RGB 6D pose estimation at category-level,\nby integrating them with feature extraction techniques. Novel View Synthesis\nallows to obtain a coarse pose that is refined through an online optimization\nmethod introduced in this work to deal with intra-category geometric\ndifferences. In such a way, the outlined method shows reduction in data\nrequirements, removal of the necessity of depth information in zero-shot\ncategory-level 6D pose estimation task, and increased performance,\nquantitatively demonstrated through experiments on the CO3D dataset.\n", "date": "2024-03-21", "categories": []}
{"id": "2307.12172", "title": "Challenges for Monocular 6D Object Pose Estimation in Robotics", "abstract": "  Object pose estimation is a core perception task that enables, for example,\nobject grasping and scene understanding. The widely available, inexpensive and\nhigh-resolution RGB sensors and CNNs that allow for fast inference based on\nthis modality make monocular approaches especially well suited for robotics\napplications. We observe that previous surveys on object pose estimation\nestablish the state of the art for varying modalities, single- and multi-view\nsettings, and datasets and metrics that consider a multitude of applications.\nWe argue, however, that those works' broad scope hinders the identification of\nopen challenges that are specific to monocular approaches and the derivation of\npromising future challenges for their application in robotics. By providing a\nunified view on recent publications from both robotics and computer vision, we\nfind that occlusion handling, novel pose representations, and formalizing and\nimproving category-level pose estimation are still fundamental challenges that\nare highly relevant for robotics. Moreover, to further improve robotic\nperformance, large object sets, novel objects, refractive materials, and\nuncertainty estimates are central, largely unsolved open challenges. In order\nto address them, ontological reasoning, deformability handling, scene-level\nreasoning, realistic datasets, and the ecological footprint of algorithms need\nto be improved.\n", "date": "2023-07-22", "categories": []}
{"id": "2310.09982", "title": "AEP$n$P: A Less-constrained EP$n$P Solver for Pose Estimation with\n  Anisotropic Scaling", "abstract": "  Perspective-$n$-Point (P$n$P) stands as a fundamental algorithm for pose\nestimation in various applications. In this paper, we present a new approach to\nthe P$n$P problem with relaxed constraints, eliminating the need for precise 3D\ncoordinates, which is especially suitable for object pose estimation where\ncorresponding object models may not be available in practice. Built upon the\nclassical EP$n$P solver, we refer to it as AEP$n$P due to its ability to handle\nunknown anisotropic scaling factors in addition to the common 6D\ntransformation. Through a few algebraic manipulations and a well-chosen frame\nof reference, this new problem can be boiled down to a simple linear null-space\nproblem followed by point registration-based identification of a similarity\ntransformation. Experimental results on both simulated and real datasets\ndemonstrate the effectiveness of AEP$n$P as a flexible and practical solution\nto object pose estimation. Code: https://github.com/goldoak/AEPnP.\n", "date": "2023-10-15", "categories": []}
{"id": "2402.04878", "title": "Shape-biased Texture Agnostic Representations for Improved Textureless\n  and Metallic Object Detection and 6D Pose Estimation", "abstract": "  Recent advances in machine learning have greatly benefited object detection\nand 6D pose estimation. However, textureless and metallic objects still pose a\nsignificant challenge due to few visual cues and the texture bias of CNNs. To\naddress his issue, we propose a strategy for inducing a shape bias to CNN\ntraining. In particular, by randomizing textures applied to object surfaces\nduring data rendering, we create training data without consistent textural\ncues. This methodology allows for seamless integration into existing data\nrendering engines, and results in negligible computational overhead for data\nrendering and network training. Our findings demonstrate that the shape bias we\ninduce via randomized texturing, improves over existing approaches using style\ntransfer. We evaluate with three detectors and two pose estimators. For the\nmost recent object detector and for pose estimation in general, estimation\naccuracy improves for textureless and metallic objects. Additionally we show\nthat our approach increases the pose estimation accuracy in the presence of\nimage noise and strong illumination changes. Code and datasets are publicly\navailable at github.com/hoenigpeter/randomized_texturing.\n", "date": "2024-02-07", "categories": []}
{"id": "2407.15484", "title": "6DGS: 6D Pose Estimation from a Single Image and a 3D Gaussian Splatting\n  Model", "abstract": "  We propose 6DGS to estimate the camera pose of a target RGB image given a 3D\nGaussian Splatting (3DGS) model representing the scene. 6DGS avoids the\niterative process typical of analysis-by-synthesis methods (e.g. iNeRF) that\nalso require an initialization of the camera pose in order to converge.\nInstead, our method estimates a 6DoF pose by inverting the 3DGS rendering\nprocess. Starting from the object surface, we define a radiant Ellicell that\nuniformly generates rays departing from each ellipsoid that parameterize the\n3DGS model. Each Ellicell ray is associated with the rendering parameters of\neach ellipsoid, which in turn is used to obtain the best bindings between the\ntarget image pixels and the cast rays. These pixel-ray bindings are then ranked\nto select the best scoring bundle of rays, which their intersection provides\nthe camera center and, in turn, the camera rotation. The proposed solution\nobviates the necessity of an \"a priori\" pose for initialization, and it solves\n6DoF pose estimation in closed form, without the need for iterations. Moreover,\ncompared to the existing Novel View Synthesis (NVS) baselines for pose\nestimation, 6DGS can improve the overall average rotational accuracy by 12% and\ntranslation accuracy by 22% on real scenes, despite not requiring any\ninitialization pose. At the same time, our method operates near real-time,\nreaching 15fps on consumer hardware.\n", "date": "2024-07-22", "categories": []}
{"id": "2311.18809", "title": "FoundPose: Unseen Object Pose Estimation with Foundation Features", "abstract": "  We propose FoundPose, a model-based method for 6D pose estimation of unseen\nobjects from a single RGB image. The method can quickly onboard new objects\nusing their 3D models without requiring any object- or task-specific training.\nIn contrast, existing methods typically pre-train on large-scale, task-specific\ndatasets in order to generalize to new objects and to bridge the image-to-model\ndomain gap. We demonstrate that such generalization capabilities can be\nobserved in a recent vision foundation model trained in a self-supervised\nmanner. Specifically, our method estimates the object pose from image-to-model\n2D-3D correspondences, which are established by matching patch descriptors from\nthe recent DINOv2 model between the image and pre-rendered object templates. We\nfind that reliable correspondences can be established by kNN matching of patch\ndescriptors from an intermediate DINOv2 layer. Such descriptors carry stronger\npositional information than descriptors from the last layer, and we show their\nimportance when semantic information is ambiguous due to object symmetries or a\nlack of texture. To avoid establishing correspondences against all object\ntemplates, we develop an efficient template retrieval approach that integrates\nthe patch descriptors into the bag-of-words representation and can promptly\npropose a handful of similarly looking templates. Additionally, we apply\nfeaturemetric alignment to compensate for discrepancies in the 2D-3D\ncorrespondences caused by coarse patch sampling. The resulting method\nnoticeably outperforms existing RGB methods for refinement-free pose estimation\non the standard BOP benchmark with seven diverse datasets and can be seamlessly\ncombined with an existing render-and-compare refinement method to achieve\nRGB-only state-of-the-art results. Project page: evinpinar.github.io/foundpose.\n", "date": "2023-11-30", "categories": []}
{"id": "2312.15130", "title": "PACE: A Large-Scale Dataset with Pose Annotations in Cluttered\n  Environments", "abstract": "  We introduce PACE (Pose Annotations in Cluttered Environments), a large-scale\nbenchmark designed to advance the development and evaluation of pose estimation\nmethods in cluttered scenarios. PACE provides a large-scale real-world\nbenchmark for both instance-level and category-level settings. The benchmark\nconsists of 55K frames with 258K annotations across 300 videos, covering 238\nobjects from 43 categories and featuring a mix of rigid and articulated items\nin cluttered scenes. To annotate the real-world data efficiently, we develop an\ninnovative annotation system with a calibrated 3-camera setup. Additionally, we\noffer PACE-Sim, which contains 100K photo-realistic simulated frames with 2.4M\nannotations across 931 objects. We test state-of-the-art algorithms in PACE\nalong two tracks: pose estimation, and object pose tracking, revealing the\nbenchmark's challenges and research opportunities. Our benchmark code and data\nis available on https://github.com/qq456cvb/PACE.\n", "date": "2023-12-23", "categories": []}
{"id": "2407.08199", "title": "SRPose: Two-view Relative Pose Estimation with Sparse Keypoints", "abstract": "  Two-view pose estimation is essential for map-free visual relocalization and\nobject pose tracking tasks. However, traditional matching methods suffer from\ntime-consuming robust estimators, while deep learning-based pose regressors\nonly cater to camera-to-world pose estimation, lacking generalizability to\ndifferent image sizes and camera intrinsics. In this paper, we propose SRPose,\na sparse keypoint-based framework for two-view relative pose estimation in\ncamera-to-world and object-to-camera scenarios. SRPose consists of a sparse\nkeypoint detector, an intrinsic-calibration position encoder, and promptable\nprior knowledge-guided attention layers. Given two RGB images of a fixed scene\nor a moving object, SRPose estimates the relative camera or 6D object pose\ntransformation. Extensive experiments demonstrate that SRPose achieves\ncompetitive or superior performance compared to state-of-the-art methods in\nterms of accuracy and speed, showing generalizability to both scenarios. It is\nrobust to different image sizes and camera intrinsics, and can be deployed with\nlow computing resources.\n", "date": "2024-07-11", "categories": []}
{"id": "2311.09500", "title": "Pseudo-keypoint RKHS Learning for Self-supervised 6DoF Pose Estimation", "abstract": "  We address the simulation-to-real domain gap in six degree-of-freedom pose\nestimation (6DoF PE), and propose a novel self-supervised keypoint voting-based\n6DoF PE framework, effectively narrowing this gap using a learnable kernel in\nRKHS. We formulate this domain gap as a distance in high-dimensional feature\nspace, distinct from previous iterative matching methods. We propose an adapter\nnetwork, which is pre-trained on purely synthetic data with synthetic ground\ntruth poses, and which evolves the network parameters from this source\nsynthetic domain to the target real domain. Importantly, the real data training\nonly uses pseudo-poses estimated by pseudo-keypoints, and thereby requires no\nreal ground truth data annotations. Our proposed method is called RKHSPose, and\nachieves state-of-the-art performance among self-supervised methods on three\ncommonly used 6DoF PE datasets including LINEMOD (+4.2%), Occlusion LINEMOD\n(+2%), and YCB-Video (+3%). It also compares favorably to fully supervised\nmethods on all six applicable BOP core datasets, achieving within -11.3% to\n+0.2% of the top fully supervised results.\n", "date": "2023-11-16", "categories": []}
{"id": "2407.06984", "title": "Category-level Object Detection, Pose Estimation and Reconstruction from\n  Stereo Images", "abstract": "  We study the 3D object understanding task for manipulating everyday objects\nwith different material properties (diffuse, specular, transparent and mixed).\nExisting monocular and RGB-D methods suffer from scale ambiguity due to missing\nor imprecise depth measurements. We present CODERS, a one-stage approach for\nCategory-level Object Detection, pose Estimation and Reconstruction from Stereo\nimages. The base of our pipeline is an implicit stereo matching module that\ncombines stereo image features with 3D position information. Concatenating this\npresented module and the following transform-decoder architecture leads to\nend-to-end learning of multiple tasks required by robot manipulation. Our\napproach significantly outperforms all competing methods in the public TOD\ndataset. Furthermore, trained on simulated data, CODERS generalize well to\nunseen category-level object instances in real-world robot manipulation\nexperiments. Our dataset, code, and demos will be available on our project\npage.\n", "date": "2024-07-09", "categories": []}
{"id": "2401.02281", "title": "PEGASUS: Physically Enhanced Gaussian Splatting Simulation System for\n  6DoF Object Pose Dataset Generation", "abstract": "  We introduce Physically Enhanced Gaussian Splatting Simulation System\n(PEGASUS) for 6DOF object pose dataset generation, a versatile dataset\ngenerator based on 3D Gaussian Splatting.\n  Environment and object representations can be easily obtained using commodity\ncameras to reconstruct with Gaussian Splatting. <i>PEGASUS</i> allows the\ncomposition of new scenes by merging the respective underlying Gaussian\nSplatting point cloud of an environment with one or multiple objects.\nLeveraging a physics engine enables the simulation of natural object placement\nwithin a scene through interaction between meshes extracted for the objects and\nthe environment. Consequently, an extensive amount of new scenes - static or\ndynamic - can be created by combining different environments and objects. By\nrendering scenes from various perspectives, diverse data points such as RGB\nimages, depth maps, semantic masks, and 6DoF object poses can be extracted.\n  Our study demonstrates that training on data generated by PEGASUS enables\npose estimation networks to successfully transfer from synthetic data to\nreal-world data. Moreover, we introduce the Ramen dataset, comprising 30\nJapanese cup noodle items. This dataset includes spherical scans that captures\nimages from both object hemisphere and the Gaussian Splatting reconstruction,\nmaking them compatible with PEGASUS.\n", "date": "2024-01-04", "categories": []}
{"id": "2407.10762", "title": "Domain Generalization for 6D Pose Estimation Through NeRF-based Image\n  Synthesis", "abstract": "  This work introduces a novel augmentation method that increases the diversity\nof a train set to improve the generalization abilities of a 6D pose estimation\nnetwork. For this purpose, a Neural Radiance Field is trained from synthetic\nimages and exploited to generate an augmented set. Our method enriches the\ninitial set by enabling the synthesis of images with (i) unseen viewpoints,\n(ii) rich illumination conditions through appearance extrapolation, and (iii)\nrandomized textures. We validate our augmentation method on the challenging\nuse-case of spacecraft pose estimation and show that it significantly improves\nthe pose estimation generalization capabilities. On the SPEED+ dataset, our\nmethod reduces the error on the pose by 50% on both target domains.\n", "date": "2024-07-15", "categories": []}
{"id": "2407.08909", "title": "KGpose: Keypoint-Graph Driven End-to-End Multi-Object 6D Pose Estimation\n  via Point-Wise Pose Voting", "abstract": "  This letter presents KGpose, a novel end-to-end framework for 6D pose\nestimation of multiple objects. Our approach combines keypoint-based method\nwith learnable pose regression through `keypoint-graph', which is a graph\nrepresentation of the keypoints. KGpose first estimates 3D keypoints for each\nobject using an attentional multi-modal feature fusion of RGB and point cloud\nfeatures. These keypoints are estimated from each point of point cloud and\nconverted into a graph representation. The network directly regresses 6D pose\nparameters for each point through a sequence of keypoint-graph embedding and\nlocal graph embedding which are designed with graph convolutions, followed by\nrotation and translation heads. The final pose for each object is selected from\nthe candidates of point-wise predictions. The method achieves competitive\nresults on the benchmark dataset, demonstrating the effectiveness of our model.\nKGpose enables multi-object pose estimation without requiring an extra\nlocalization step, offering a unified and efficient solution for understanding\ngeometric contexts in complex scenes for robotic applications.\n", "date": "2024-07-12", "categories": []}
{"id": "2406.16384", "title": "High-resolution open-vocabulary object 6D pose estimation", "abstract": "  The generalisation to unseen objects in the 6D pose estimation task is very\nchallenging. While Vision-Language Models (VLMs) enable using natural language\ndescriptions to support 6D pose estimation of unseen objects, these solutions\nunderperform compared to model-based methods. In this work we present Horyon,\nan open-vocabulary VLM-based architecture that addresses relative pose\nestimation between two scenes of an unseen object, described by a textual\nprompt only. We use the textual prompt to identify the unseen object in the\nscenes and then obtain high-resolution multi-scale features. These features are\nused to extract cross-scene matches for registration. We evaluate our model on\na benchmark with a large variety of unseen objects across four datasets, namely\nREAL275, Toyota-Light, Linemod, and YCB-Video. Our method achieves\nstate-of-the-art performance on all datasets, outperforming by 12.6 in Average\nRecall the previous best-performing approach.\n", "date": "2024-06-24", "categories": []}
{"id": "2407.08711", "title": "OmniNOCS: A unified NOCS dataset and model for 3D lifting of 2D objects", "abstract": "  We propose OmniNOCS, a large-scale monocular dataset with 3D Normalized\nObject Coordinate Space (NOCS) maps, object masks, and 3D bounding box\nannotations for indoor and outdoor scenes. OmniNOCS has 20 times more object\nclasses and 200 times more instances than existing NOCS datasets (NOCS-Real275,\nWild6D). We use OmniNOCS to train a novel, transformer-based monocular NOCS\nprediction model (NOCSformer) that can predict accurate NOCS, instance masks\nand poses from 2D object detections across diverse classes. It is the first\nNOCS model that can generalize to a broad range of classes when prompted with\n2D boxes. We evaluate our model on the task of 3D oriented bounding box\nprediction, where it achieves comparable results to state-of-the-art 3D\ndetection methods such as Cube R-CNN. Unlike other 3D detection methods, our\nmodel also provides detailed and accurate 3D object shape and segmentation. We\npropose a novel benchmark for the task of NOCS prediction based on OmniNOCS,\nwhich we hope will serve as a useful baseline for future work in this area. Our\ndataset and code will be at the project website: https://omninocs.github.io.\n", "date": "2024-07-11", "categories": []}
{"id": "2407.04384", "title": "Unsupervised Learning of Category-Level 3D Pose from Object-Centric\n  Videos", "abstract": "  Category-level 3D pose estimation is a fundamentally important problem in\ncomputer vision and robotics, e.g. for embodied agents or to train 3D\ngenerative models. However, so far methods that estimate the category-level\nobject pose require either large amounts of human annotations, CAD models or\ninput from RGB-D sensors. In contrast, we tackle the problem of learning to\nestimate the category-level 3D pose only from casually taken object-centric\nvideos without human supervision. We propose a two-step pipeline: First, we\nintroduce a multi-view alignment procedure that determines canonical camera\nposes across videos with a novel and robust cyclic distance formulation for\ngeometric and appearance matching using reconstructed coarse meshes and DINOv2\nfeatures. In a second step, the canonical poses and reconstructed meshes enable\nus to train a model for 3D pose estimation from a single image. In particular,\nour model learns to estimate dense correspondences between images and a\nprototypical 3D template by predicting, for each pixel in a 2D image, a feature\nvector of the corresponding vertex in the template mesh. We demonstrate that\nour method outperforms all baselines at the unsupervised alignment of\nobject-centric videos by a large margin and provides faithful and robust\npredictions in-the-wild. Our code and data is available at\nhttps://github.com/GenIntel/uns-obj-pose3d.\n", "date": "2024-07-05", "categories": []}
{"id": "2312.00690", "title": "Open-vocabulary object 6D pose estimation", "abstract": "  We introduce the new setting of open-vocabulary object 6D pose estimation, in\nwhich a textual prompt is used to specify the object of interest. In contrast\nto existing approaches, in our setting (i) the object of interest is specified\nsolely through the textual prompt, (ii) no object model (e.g., CAD or video\nsequence) is required at inference, and (iii) the object is imaged from two\nRGBD viewpoints of different scenes. To operate in this setting, we introduce a\nnovel approach that leverages a Vision-Language Model to segment the object of\ninterest from the scenes and to estimate its relative 6D pose. The key of our\napproach is a carefully devised strategy to fuse object-level information\nprovided by the prompt with local image features, resulting in a feature space\nthat can generalize to novel concepts. We validate our approach on a new\nbenchmark based on two popular datasets, REAL275 and Toyota-Light, which\ncollectively encompass 34 object instances appearing in four thousand image\npairs. The results demonstrate that our approach outperforms both a\nwell-established hand-crafted method and a recent deep learning-based baseline\nin estimating the relative 6D pose of objects in different scenes. Code and\ndataset are available at https://jcorsetti.github.io/oryon.\n", "date": "2023-12-01", "categories": []}
{"id": "2406.13796", "title": "NeRF-Feat: 6D Object Pose Estimation using Feature Rendering", "abstract": "  Object Pose Estimation is a crucial component in robotic grasping and\naugmented reality. Learning based approaches typically require training data\nfrom a highly accurate CAD model or labeled training data acquired using a\ncomplex setup. We address this by learning to estimate pose from weakly labeled\ndata without a known CAD model. We propose to use a NeRF to learn object shape\nimplicitly which is later used to learn view-invariant features in conjunction\nwith CNN using a contrastive loss. While NeRF helps in learning features that\nare view-consistent, CNN ensures that the learned features respect symmetry.\nDuring inference, CNN is used to predict view-invariant features which can be\nused to establish correspondences with the implicit 3d model in NeRF. The\ncorrespondences are then used to estimate the pose in the reference frame of\nNeRF. Our approach can also handle symmetric objects unlike other approaches\nusing a similar training setup. Specifically, we learn viewpoint invariant,\ndiscriminative features using NeRF which are later used for pose estimation. We\nevaluated our approach on LM, LM-Occlusion, and T-Less dataset and achieved\nbenchmark accuracy despite using weakly labeled data.\n", "date": "2024-06-19", "categories": []}
{"id": "2406.11063", "title": "FastPoseCNN: Real-Time Monocular Category-Level Pose and Size Estimation\n  Framework", "abstract": "  The primary focus of this paper is the development of a framework for pose\nand size estimation of unseen objects given a single RGB image - all in\nreal-time. In 2019, the first category-level pose and size estimation framework\nwas proposed alongside two novel datasets called CAMERA and REAL. However,\ncurrent methodologies are restricted from practical use because of its long\ninference time (2-4 fps). Their approach's inference had significant delays\nbecause they used the computationally expensive MaskedRCNN framework and\nUmeyama algorithm. To optimize our method and yield real-time results, our\nframework uses the efficient ResNet-FPN framework alongside decoupling the\ntranslation, rotation, and size regression problem by using distinct decoders.\nMoreover, our methodology performs pose and size estimation in a global context\n- i.e., estimating the involved parameters of all captured objects in the image\nall at once. We perform extensive testing to fully compare the performance in\nterms of precision and speed to demonstrate the capability of our method.\n", "date": "2024-06-16", "categories": []}
{"id": "2406.09613", "title": "ImageNet3D: Towards General-Purpose Object-Level 3D Understanding", "abstract": "  A vision model with general-purpose object-level 3D understanding should be\ncapable of inferring both 2D (e.g., class name and bounding box) and 3D\ninformation (e.g., 3D location and 3D viewpoint) for arbitrary rigid objects in\nnatural images. This is a challenging task, as it involves inferring 3D\ninformation from 2D signals and most importantly, generalizing to rigid objects\nfrom unseen categories. However, existing datasets with object-level 3D\nannotations are often limited by the number of categories or the quality of\nannotations. Models developed on these datasets become specialists for certain\ncategories or domains, and fail to generalize. In this work, we present\nImageNet3D, a large dataset for general-purpose object-level 3D understanding.\nImageNet3D augments 200 categories from the ImageNet dataset with 2D bounding\nbox, 3D pose, 3D location annotations, and image captions interleaved with 3D\ninformation. With the new annotations available in ImageNet3D, we could (i)\nanalyze the object-level 3D awareness of visual foundation models, and (ii)\nstudy and develop general-purpose models that infer both 2D and 3D information\nfor arbitrary rigid objects in natural images, and (iii) integrate unified 3D\nmodels with large language models for 3D-related reasoning.. We consider two\nnew tasks, probing of object-level 3D awareness and open vocabulary pose\nestimation, besides standard classification and pose estimation. Experimental\nresults on ImageNet3D demonstrate the potential of our dataset in building\nvision models with stronger general-purpose object-level 3D understanding.\n", "date": "2024-06-13", "categories": []}
{"id": "2405.12728", "title": "Leveraging Neural Radiance Fields for Pose Estimation of an Unknown\n  Space Object during Proximity Operations", "abstract": "  We address the estimation of the 6D pose of an unknown target spacecraft\nrelative to a monocular camera, a key step towards the autonomous rendezvous\nand proximity operations required by future Active Debris Removal missions. We\npresent a novel method that enables an \"off-the-shelf\" spacecraft pose\nestimator, which is supposed to known the target CAD model, to be applied on an\nunknown target. Our method relies on an in-the wild NeRF, i.e., a Neural\nRadiance Field that employs learnable appearance embeddings to represent\nvarying illumination conditions found in natural scenes. We train the NeRF\nmodel using a sparse collection of images that depict the target, and in turn\ngenerate a large dataset that is diverse both in terms of viewpoint and\nillumination. This dataset is then used to train the pose estimation network.\nWe validate our method on the Hardware-In-the-Loop images of SPEED+ that\nemulate lighting conditions close to those encountered on orbit. We demonstrate\nthat our method successfully enables the training of an off-the-shelf\nspacecraft pose estimation network from a sparse set of images. Furthermore, we\nshow that a network trained using our method performs similarly to a model\ntrained on synthetic images generated using the CAD model of the target.\n", "date": "2024-05-21", "categories": []}
{"id": "2406.04316", "title": "Omni6DPose: A Benchmark and Model for Universal 6D Object Pose\n  Estimation and Tracking", "abstract": "  6D Object Pose Estimation is a crucial yet challenging task in computer\nvision, suffering from a significant lack of large-scale datasets. This\nscarcity impedes comprehensive evaluation of model performance, limiting\nresearch advancements. Furthermore, the restricted number of available\ninstances or categories curtails its applications. To address these issues,\nthis paper introduces Omni6DPose, a substantial dataset characterized by its\ndiversity in object categories, large scale, and variety in object materials.\nOmni6DPose is divided into three main components: ROPE (Real 6D Object Pose\nEstimation Dataset), which includes 332K images annotated with over 1.5M\nannotations across 581 instances in 149 categories; SOPE(Simulated 6D Object\nPose Estimation Dataset), consisting of 475K images created in a mixed reality\nsetting with depth simulation, annotated with over 5M annotations across 4162\ninstances in the same 149 categories; and the manually aligned real scanned\nobjects used in both ROPE and SOPE. Omni6DPose is inherently challenging due to\nthe substantial variations and ambiguities. To address this challenge, we\nintroduce GenPose++, an enhanced version of the SOTA category-level pose\nestimation framework, incorporating two pivotal improvements: Semantic-aware\nfeature extraction and Clustering-based aggregation. Moreover, we provide a\ncomprehensive benchmarking analysis to evaluate the performance of previous\nmethods on this large-scale dataset in the realms of 6D object pose estimation\nand pose tracking.\n", "date": "2024-06-06", "categories": []}
{"id": "2406.02977", "title": "Sparse Color-Code Net: Real-Time RGB-Based 6D Object Pose Estimation on\n  Edge Devices", "abstract": "  As robotics and augmented reality applications increasingly rely on precise\nand efficient 6D object pose estimation, real-time performance on edge devices\nis required for more interactive and responsive systems. Our proposed Sparse\nColor-Code Net (SCCN) embodies a clear and concise pipeline design to\neffectively address this requirement. SCCN performs pixel-level predictions on\nthe target object in the RGB image, utilizing the sparsity of essential object\ngeometry features to speed up the Perspective-n-Point (PnP) computation\nprocess. Additionally, it introduces a novel pixel-level geometry-based object\nsymmetry representation that seamlessly integrates with the initial pose\npredictions, effectively addressing symmetric object ambiguities. SCCN notably\nachieves an estimation rate of 19 frames per second (FPS) and 6 FPS on the\nbenchmark LINEMOD dataset and the Occlusion LINEMOD dataset, respectively, for\nan NVIDIA Jetson AGX Xavier, while consistently maintaining high estimation\naccuracy at these rates.\n", "date": "2024-06-05", "categories": []}
{"id": "2311.18107", "title": "A Stochastic-Geometrical Framework for Object Pose Estimation based on\n  Mixture Models Avoiding the Correspondence Problem", "abstract": "  Background: Pose estimation of rigid objects is a practical challenge in\noptical metrology and computer vision. This paper presents a novel\nstochastic-geometrical modeling framework for object pose estimation based on\nobserving multiple feature points.\n  Methods: This framework utilizes mixture models for feature point densities\nin object space and for interpreting real measurements. Advantages are the\navoidance to resolve individual feature correspondences and to incorporate\ncorrect stochastic dependencies in multi-view applications. First, the general\nmodeling framework is presented, second, a general algorithm for pose\nestimation is derived, and third, two example models (camera and lateration\nsetup) are presented.\n  Results: Numerical experiments show the effectiveness of this modeling and\ngeneral algorithm by presenting four simulation scenarios for three observation\nsystems, including the dependence on measurement resolution, object\ndeformations and measurement noise. Probabilistic modeling utilizing mixture\nmodels shows the potential for accurate and robust pose estimations while\navoiding the correspondence problem.\n", "date": "2023-11-29", "categories": []}
{"id": "2405.10557", "title": "Resolving Symmetry Ambiguity in Correspondence-based Methods for\n  Instance-level Object Pose Estimation", "abstract": "  Estimating the 6D pose of an object from a single RGB image is a critical\ntask that becomes additionally challenging when dealing with symmetric objects.\nRecent approaches typically establish one-to-one correspondences between image\npixels and 3D object surface vertices. However, the utilization of one-to-one\ncorrespondences introduces ambiguity for symmetric objects. To address this, we\npropose SymCode, a symmetry-aware surface encoding that encodes the object\nsurface vertices based on one-to-many correspondences, eliminating the problem\nof one-to-one correspondence ambiguity. We also introduce SymNet, a fast\nend-to-end network that directly regresses the 6D pose parameters without\nsolving a PnP problem. We demonstrate faster runtime and comparable accuracy\nachieved by our method on the T-LESS and IC-BIN benchmarks of mostly symmetric\nobjects. Our source code will be released upon acceptance.\n", "date": "2024-05-17", "categories": []}
{"id": "2405.08483", "title": "RDPN6D: Residual-based Dense Point-wise Network for 6Dof Object Pose\n  Estimation Based on RGB-D Images", "abstract": "  In this work, we introduce a novel method for calculating the 6DoF pose of an\nobject using a single RGB-D image. Unlike existing methods that either directly\npredict objects' poses or rely on sparse keypoints for pose recovery, our\napproach addresses this challenging task using dense correspondence, i.e., we\nregress the object coordinates for each visible pixel. Our method leverages\nexisting object detection methods. We incorporate a re-projection mechanism to\nadjust the camera's intrinsic matrix to accommodate cropping in RGB-D images.\nMoreover, we transform the 3D object coordinates into a residual\nrepresentation, which can effectively reduce the output space and yield\nsuperior performance. We conducted extensive experiments to validate the\nefficacy of our approach for 6D pose estimation. Our approach outperforms most\nprevious methods, especially in occlusion scenarios, and demonstrates notable\nimprovements over the state-of-the-art methods. Our code is available on\nhttps://github.com/AI-Application-and-Integration-Lab/RDPN6D.\n", "date": "2024-05-14", "categories": []}
{"id": "2403.01517", "title": "MatchU: Matching Unseen Objects for 6D Pose Estimation from RGB-D Images", "abstract": "  Recent learning methods for object pose estimation require resource-intensive\ntraining for each individual object instance or category, hampering their\nscalability in real applications when confronted with previously unseen\nobjects. In this paper, we propose MatchU, a Fuse-Describe-Match strategy for\n6D pose estimation from RGB-D images. MatchU is a generic approach that fuses\n2D texture and 3D geometric cues for 6D pose prediction of unseen objects. We\nrely on learning geometric 3D descriptors that are rotation-invariant by\ndesign. By encoding pose-agnostic geometry, the learned descriptors naturally\ngeneralize to unseen objects and capture symmetries. To tackle ambiguous\nassociations using 3D geometry only, we fuse additional RGB information into\nour descriptor. This is achieved through a novel attention-based mechanism that\nfuses cross-modal information, together with a matching loss that leverages the\nlatent space learned from RGB data to guide the descriptor learning process.\nExtensive experiments reveal the generalizability of both the RGB-D fusion\nstrategy as well as the descriptor efficacy. Benefiting from the novel designs,\nMatchU surpasses all existing methods by a significant margin in terms of both\naccuracy and speed, even without the requirement of expensive re-training or\nrendering.\n", "date": "2024-03-03", "categories": []}
{"id": "2402.15726", "title": "CLIPose: Category-Level Object Pose Estimation with Pre-trained\n  Vision-Language Knowledge", "abstract": "  Most of existing category-level object pose estimation methods devote to\nlearning the object category information from point cloud modality. However,\nthe scale of 3D datasets is limited due to the high cost of 3D data collection\nand annotation. Consequently, the category features extracted from these\nlimited point cloud samples may not be comprehensive. This motivates us to\ninvestigate whether we can draw on knowledge of other modalities to obtain\ncategory information. Inspired by this motivation, we propose CLIPose, a novel\n6D pose framework that employs the pre-trained vision-language model to develop\nbetter learning of object category information, which can fully leverage\nabundant semantic knowledge in image and text modalities. To make the 3D\nencoder learn category-specific features more efficiently, we align\nrepresentations of three modalities in feature space via multi-modal\ncontrastive learning. In addition to exploiting the pre-trained knowledge of\nthe CLIP's model, we also expect it to be more sensitive with pose parameters.\nTherefore, we introduce a prompt tuning approach to fine-tune image encoder\nwhile we incorporate rotations and translations information in the text\ndescriptions. CLIPose achieves state-of-the-art performance on two mainstream\nbenchmark datasets, REAL275 and CAMERA25, and runs in real-time during\ninference (40FPS).\n", "date": "2024-02-24", "categories": []}
{"id": "2404.18381", "title": "Object Registration in Neural Fields", "abstract": "  Neural fields provide a continuous scene representation of 3D geometry and\nappearance in a way which has great promise for robotics applications. One\nfunctionality that unlocks unique use-cases for neural fields in robotics is\nobject 6-DoF registration. In this paper, we provide an expanded analysis of\nthe recent Reg-NF neural field registration method and its use-cases within a\nrobotics context. We showcase the scenario of determining the 6-DoF pose of\nknown objects within a scene using scene and object neural field models. We\nshow how this may be used to better represent objects within imperfectly\nmodelled scenes and generate new scenes by substituting object neural field\nmodels into the scene.\n", "date": "2024-04-29", "categories": []}
{"id": "2403.07741", "title": "Uncertainty Quantification with Deep Ensembles for 6D Object Pose\n  Estimation", "abstract": "  The estimation of 6D object poses is a fundamental task in many computer\nvision applications. Particularly, in high risk scenarios such as human-robot\ninteraction, industrial inspection, and automation, reliable pose estimates are\ncrucial. In the last years, increasingly accurate and robust\ndeep-learning-based approaches for 6D object pose estimation have been\nproposed. Many top-performing methods are not end-to-end trainable but consist\nof multiple stages. In the context of deep uncertainty quantification, deep\nensembles are considered as state of the art since they have been proven to\nproduce well-calibrated and robust uncertainty estimates. However, deep\nensembles can only be applied to methods that can be trained end-to-end. In\nthis work, we propose a method to quantify the uncertainty of multi-stage 6D\nobject pose estimation approaches with deep ensembles. For the implementation,\nwe choose SurfEmb as representative, since it is one of the top-performing 6D\nobject pose estimation approaches in the BOP Challenge 2022. We apply\nestablished metrics and concepts for deep uncertainty quantification to\nevaluate the results. Furthermore, we propose a novel uncertainty calibration\nscore for regression tasks to quantify the quality of the estimated\nuncertainty.\n", "date": "2024-03-12", "categories": []}
{"id": "2404.16558", "title": "DeepKalPose: An Enhanced Deep-Learning Kalman Filter for Temporally\n  Consistent Monocular Vehicle Pose Estimation", "abstract": "  This paper presents DeepKalPose, a novel approach for enhancing temporal\nconsistency in monocular vehicle pose estimation applied on video through a\ndeep-learning-based Kalman Filter. By integrating a Bi-directional Kalman\nfilter strategy utilizing forward and backward time-series processing, combined\nwith a learnable motion model to represent complex motion patterns, our method\nsignificantly improves pose accuracy and robustness across various conditions,\nparticularly for occluded or distant vehicles. Experimental validation on the\nKITTI dataset confirms that DeepKalPose outperforms existing methods in both\npose accuracy and temporal consistency.\n", "date": "2024-04-25", "categories": []}
{"id": "2310.16279", "title": "TransPose: 6D Object Pose Estimation with Geometry-Aware Transformer", "abstract": "  Estimating the 6D object pose is an essential task in many applications. Due\nto the lack of depth information, existing RGB-based methods are sensitive to\nocclusion and illumination changes. How to extract and utilize the geometry\nfeatures in depth information is crucial to achieve accurate predictions. To\nthis end, we propose TransPose, a novel 6D pose framework that exploits\nTransformer Encoder with geometry-aware module to develop better learning of\npoint cloud feature representations. Specifically, we first uniformly sample\npoint cloud and extract local geometry features with the designed local feature\nextractor base on graph convolution network. To improve robustness to\nocclusion, we adopt Transformer to perform the exchange of global information,\nmaking each local feature contains global information. Finally, we introduce\ngeometry-aware module in Transformer Encoder, which to form an effective\nconstrain for point cloud feature learning and makes the global information\nexchange more tightly coupled with point cloud tasks. Extensive experiments\nindicate the effectiveness of TransPose, our pose estimation pipeline achieves\ncompetitive results on three benchmark datasets.\n", "date": "2023-10-25", "categories": []}
{"id": "2404.12144", "title": "Mushroom Segmentation and 3D Pose Estimation from Point Clouds using\n  Fully Convolutional Geometric Features and Implicit Pose Encoding", "abstract": "  Modern agricultural applications rely more and more on deep learning\nsolutions. However, training well-performing deep networks requires a large\namount of annotated data that may not be available and in the case of 3D\nannotation may not even be feasible for human annotators. In this work, we\ndevelop a deep learning approach to segment mushrooms and estimate their pose\non 3D data, in the form of point clouds acquired by depth sensors. To\ncircumvent the annotation problem, we create a synthetic dataset of mushroom\nscenes, where we are fully aware of 3D information, such as the pose of each\nmushroom. The proposed network has a fully convolutional backbone, that parses\nsparse 3D data, and predicts pose information that implicitly defines both\ninstance segmentation and pose estimation task. We have validated the\neffectiveness of the proposed implicit-based approach for a synthetic test set,\nas well as provided qualitative results for a small set of real acquired point\nclouds with depth sensors. Code is publicly available at\nhttps://github.com/georgeretsi/mushroom-pose.\n", "date": "2024-04-17", "categories": []}
{"id": "2403.09799", "title": "BOP Challenge 2023 on Detection, Segmentation and Pose Estimation of\n  Seen and Unseen Rigid Objects", "abstract": "  We present the evaluation methodology, datasets and results of the BOP\nChallenge 2023, the fifth in a series of public competitions organized to\ncapture the state of the art in model-based 6D object pose estimation from an\nRGB/RGB-D image and related tasks. Besides the three tasks from 2022\n(model-based 2D detection, 2D segmentation, and 6D localization of objects seen\nduring training), the 2023 challenge introduced new variants of these tasks\nfocused on objects unseen during training. In the new tasks, methods were\nrequired to learn new objects during a short onboarding stage (max 5 minutes, 1\nGPU) from provided 3D object models. The best 2023 method for 6D localization\nof unseen objects (GenFlow) notably reached the accuracy of the best 2020\nmethod for seen objects (CosyPose), although being noticeably slower. The best\n2023 method for seen objects (GPose) achieved a moderate accuracy improvement\nbut a significant 43% run-time improvement compared to the best 2022\ncounterpart (GDRNPP). Since 2017, the accuracy of 6D localization of seen\nobjects has improved by more than 50% (from 56.9 to 85.6 AR_C). The online\nevaluation system stays open and is available at: http://bop.felk.cvut.cz/.\n", "date": "2024-03-14", "categories": []}
{"id": "2404.11139", "title": "GeoReF: Geometric Alignment Across Shape Variation for Category-level\n  Object Pose Refinement", "abstract": "  Object pose refinement is essential for robust object pose estimation.\nPrevious work has made significant progress towards instance-level object pose\nrefinement. Yet, category-level pose refinement is a more challenging problem\ndue to large shape variations within a category and the discrepancies between\nthe target object and the shape prior. To address these challenges, we\nintroduce a novel architecture for category-level object pose refinement. Our\napproach integrates an HS-layer and learnable affine transformations, which\naims to enhance the extraction and alignment of geometric information.\nAdditionally, we introduce a cross-cloud transformation mechanism that\nefficiently merges diverse data sources. Finally, we push the limits of our\nmodel by incorporating the shape prior information for translation and size\nerror prediction. We conducted extensive experiments to demonstrate the\neffectiveness of the proposed framework. Through extensive quantitative\nexperiments, we demonstrate significant improvement over the baseline method by\na large margin across all metrics.\n", "date": "2024-04-17", "categories": []}
{"id": "2309.14265", "title": "Industrial Application of 6D Pose Estimation for Robotic Manipulation in\n  Automotive Internal Logistics", "abstract": "  Despite the advances in robotics a large proportion of the of parts handling\ntasks in the automotive industry's internal logistics are not automated but\nstill performed by humans. A key component to competitively automate these\nprocesses is a 6D pose estimation that can handle a large number of different\nparts, is adaptable to new parts with little manual effort, and is sufficiently\naccurate and robust with respect to industry requirements. In this context, the\nquestion arises as to the current status quo with respect to these measures. To\naddress this we built a representative 6D pose estimation pipeline with\nstate-of-the-art components from economically scalable real to synthetic data\ngeneration to pose estimators and evaluated it on automotive parts with regards\nto a realistic sequencing process. We found that using the data generation\napproaches, the performance of the trained 6D pose estimators are promising,\nbut do not meet industry requirements. We reveal that the reason for this is\nthe inability of the estimators to provide reliable uncertainties for their\nposes, rather than the ability of to provide sufficiently accurate poses. In\nthis context we further analyzed how RGB- and RGB-D-based approaches compare\nagainst this background and show that they are differently vulnerable to the\ndomain gap induced by synthetic data.\n", "date": "2023-09-25", "categories": []}
{"id": "2305.15873", "title": "Confronting Ambiguity in 6D Object Pose Estimation via Score-Based\n  Diffusion on SE(3)", "abstract": "  Addressing pose ambiguity in 6D object pose estimation from single RGB images\npresents a significant challenge, particularly due to object symmetries or\nocclusions. In response, we introduce a novel score-based diffusion method\napplied to the $SE(3)$ group, marking the first application of diffusion models\nto $SE(3)$ within the image domain, specifically tailored for pose estimation\ntasks. Extensive evaluations demonstrate the method's efficacy in handling pose\nambiguity, mitigating perspective-induced ambiguity, and showcasing the\nrobustness of our surrogate Stein score formulation on $SE(3)$. This\nformulation not only improves the convergence of denoising process but also\nenhances computational efficiency. Thus, we pioneer a promising strategy for 6D\nobject pose estimation.\n", "date": "2023-05-25", "categories": []}
{"id": "2311.12588", "title": "HiPose: Hierarchical Binary Surface Encoding and Correspondence Pruning\n  for RGB-D 6DoF Object Pose Estimation", "abstract": "  In this work, we present a novel dense-correspondence method for 6DoF object\npose estimation from a single RGB-D image. While many existing data-driven\nmethods achieve impressive performance, they tend to be time-consuming due to\ntheir reliance on rendering-based refinement approaches. To circumvent this\nlimitation, we present HiPose, which establishes 3D-3D correspondences in a\ncoarse-to-fine manner with a hierarchical binary surface encoding. Unlike\nprevious dense-correspondence methods, we estimate the correspondence surface\nby employing point-to-surface matching and iteratively constricting the surface\nuntil it becomes a correspondence point while gradually removing outliers.\nExtensive experiments on public benchmarks LM-O, YCB-V, and T-Less demonstrate\nthat our method surpasses all refinement-free methods and is even on par with\nexpensive refinement-based approaches. Crucially, our approach is\ncomputationally efficient and enables real-time critical applications with high\naccuracy requirements.\n", "date": "2023-11-21", "categories": []}
{"id": "2404.05626", "title": "Learning a Category-level Object Pose Estimator without Pose Annotations", "abstract": "  3D object pose estimation is a challenging task. Previous works always\nrequire thousands of object images with annotated poses for learning the 3D\npose correspondence, which is laborious and time-consuming for labeling. In\nthis paper, we propose to learn a category-level 3D object pose estimator\nwithout pose annotations. Instead of using manually annotated images, we\nleverage diffusion models (e.g., Zero-1-to-3) to generate a set of images under\ncontrolled pose differences and propose to learn our object pose estimator with\nthose images. Directly using the original diffusion model leads to images with\nnoisy poses and artifacts. To tackle this issue, firstly, we exploit an image\nencoder, which is learned from a specially designed contrastive pose learning,\nto filter the unreasonable details and extract image feature maps.\nAdditionally, we propose a novel learning strategy that allows the model to\nlearn object poses from those generated image sets without knowing the\nalignment of their canonical poses. Experimental results show that our method\nhas the capability of category-level object pose estimation from a single shot\nsetting (as pose definition), while significantly outperforming other\nstate-of-the-art methods on the few-shot category-level object pose estimation\nbenchmarks.\n", "date": "2024-04-08", "categories": []}
{"id": "2303.13612", "title": "NOPE: Novel Object Pose Estimation from a Single Image", "abstract": "  The practicality of 3D object pose estimation remains limited for many\napplications due to the need for prior knowledge of a 3D model and a training\nperiod for new objects. To address this limitation, we propose an approach that\ntakes a single image of a new object as input and predicts the relative pose of\nthis object in new images without prior knowledge of the object's 3D model and\nwithout requiring training time for new objects and categories. We achieve this\nby training a model to directly predict discriminative embeddings for\nviewpoints surrounding the object. This prediction is done using a simple U-Net\narchitecture with attention and conditioned on the desired pose, which yields\nextremely fast inference. We compare our approach to state-of-the-art methods\nand show it outperforms them both in terms of accuracy and robustness. Our\nsource code is publicly available at https://github.com/nv-nguyen/nope\n", "date": "2023-03-23", "categories": []}
{"id": "2404.00891", "title": "Marrying NeRF with Feature Matching for One-step Pose Estimation", "abstract": "  Given the image collection of an object, we aim at building a real-time\nimage-based pose estimation method, which requires neither its CAD model nor\nhours of object-specific training. Recent NeRF-based methods provide a\npromising solution by directly optimizing the pose from pixel loss between\nrendered and target images. However, during inference, they require long\nconverging time, and suffer from local minima, making them impractical for\nreal-time robot applications. We aim at solving this problem by marrying image\nmatching with NeRF. With 2D matches and depth rendered by NeRF, we directly\nsolve the pose in one step by building 2D-3D correspondences between target and\ninitial view, thus allowing for real-time prediction. Moreover, to improve the\naccuracy of 2D-3D correspondences, we propose a 3D consistent point mining\nstrategy, which effectively discards unfaithful points reconstruted by NeRF.\nMoreover, current NeRF-based methods naively optimizing pixel loss fail at\noccluded images. Thus, we further propose a 2D matches based sampling strategy\nto preclude the occluded area. Experimental results on representative datasets\nprove that our method outperforms state-of-the-art methods, and improves\ninference efficiency by 90x, achieving real-time prediction at 6 FPS.\n", "date": "2024-04-01", "categories": []}
{"id": "2211.13398", "title": "CPPF++: Uncertainty-Aware Sim2Real Object Pose Estimation by Vote\n  Aggregation", "abstract": "  Object pose estimation constitutes a critical area within the domain of 3D\nvision. While contemporary state-of-the-art methods that leverage real-world\npose annotations have demonstrated commendable performance, the procurement of\nsuch real training data incurs substantial costs. This paper focuses on a\nspecific setting wherein only 3D CAD models are utilized as a priori knowledge,\ndevoid of any background or clutter information. We introduce a novel method,\nCPPF++, designed for sim-to-real pose estimation. This method builds upon the\nfoundational point-pair voting scheme of CPPF, reformulating it through a\nprobabilistic view. To address the challenge posed by vote collision, we\npropose a novel approach that involves modeling the voting uncertainty by\nestimating the probabilistic distribution of each point pair within the\ncanonical space. Furthermore, we augment the contextual information provided by\neach voting unit through the introduction of N-point tuples. To enhance the\nrobustness and accuracy of the model, we incorporate several innovative\nmodules, including noisy pair filtering, online alignment optimization, and a\ntuple feature ensemble. Alongside these methodological advancements, we\nintroduce a new category-level pose estimation dataset, named DiversePose 300.\nEmpirical evidence demonstrates that our method significantly surpasses\nprevious sim-to-real approaches and achieves comparable or superior performance\non novel datasets. Our code is available on https://github.com/qq456cvb/CPPF2.\n", "date": "2022-11-24", "categories": []}
{"id": "2403.19527", "title": "Instance-Adaptive and Geometric-Aware Keypoint Learning for\n  Category-Level 6D Object Pose Estimation", "abstract": "  Category-level 6D object pose estimation aims to estimate the rotation,\ntranslation and size of unseen instances within specific categories. In this\narea, dense correspondence-based methods have achieved leading performance.\nHowever, they do not explicitly consider the local and global geometric\ninformation of different instances, resulting in poor generalization ability to\nunseen instances with significant shape variations. To deal with this problem,\nwe propose a novel Instance-Adaptive and Geometric-Aware Keypoint Learning\nmethod for category-level 6D object pose estimation (AG-Pose), which includes\ntwo key designs: (1) The first design is an Instance-Adaptive Keypoint\nDetection module, which can adaptively detect a set of sparse keypoints for\nvarious instances to represent their geometric structures. (2) The second\ndesign is a Geometric-Aware Feature Aggregation module, which can efficiently\nintegrate the local and global geometric information into keypoint features.\nThese two modules can work together to establish robust keypoint-level\ncorrespondences for unseen instances, thus enhancing the generalization ability\nof the model.Experimental results on CAMERA25 and REAL275 datasets show that\nthe proposed AG-Pose outperforms state-of-the-art methods by a large margin\nwithout category-specific shape priors.\n", "date": "2024-03-28", "categories": []}
{"id": "2312.08344", "title": "FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects", "abstract": "  We present FoundationPose, a unified foundation model for 6D object pose\nestimation and tracking, supporting both model-based and model-free setups. Our\napproach can be instantly applied at test-time to a novel object without\nfine-tuning, as long as its CAD model is given, or a small number of reference\nimages are captured. We bridge the gap between these two setups with a neural\nimplicit representation that allows for effective novel view synthesis, keeping\nthe downstream pose estimation modules invariant under the same unified\nframework. Strong generalizability is achieved via large-scale synthetic\ntraining, aided by a large language model (LLM), a novel transformer-based\narchitecture, and contrastive learning formulation. Extensive evaluation on\nmultiple public datasets involving challenging scenarios and objects indicate\nour unified approach outperforms existing methods specialized for each task by\na large margin. In addition, it even achieves comparable results to\ninstance-level methods despite the reduced assumptions. Project page:\nhttps://nvlabs.github.io/FoundationPose/\n", "date": "2023-12-13", "categories": []}
{"id": "2308.08856", "title": "MV-ROPE: Multi-view Constraints for Robust Category-level Object Pose\n  and Size Estimation", "abstract": "  Recently there has been a growing interest in category-level object pose and\nsize estimation, and prevailing methods commonly rely on single view RGB-D\nimages. However, one disadvantage of such methods is that they require accurate\ndepth maps which cannot be produced by consumer-grade sensors. Furthermore,\nmany practical real-world situations involve a moving camera that continuously\nobserves its surroundings, and the temporal information of the input video\nstreams is simply overlooked by single-view methods. We propose a novel\nsolution that makes use of RGB video streams. Our framework consists of three\nmodules: a scale-aware monocular dense SLAM solution, a lightweight object pose\npredictor, and an object-level pose graph optimizer. The SLAM module utilizes a\nvideo stream and additional scale-sensitive readings to estimate camera poses\nand metric depth. The object pose predictor then generates canonical object\nrepresentations from RGB images. The object pose is estimated through geometric\nregistration of these canonical object representations with estimated object\ndepth points. All per-view estimates finally undergo optimization within a pose\ngraph, culminating in the output of robust and accurate canonical object poses.\nOur experimental results demonstrate that when utilizing public dataset\nsequences with high-quality depth information, the proposed method exhibits\ncomparable performance to state-of-the-art RGB-D methods. We also collect and\nevaluate on new datasets containing depth maps of varying quality to further\nquantitatively benchmark the proposed method alongside previous RGB-D based\nmethods. We demonstrate a significant advantage in scenarios where depth input\nis absent or the quality of depth sensing is limited.\n", "date": "2023-08-17", "categories": []}
{"id": "2311.11125", "title": "SecondPose: SE(3)-Consistent Dual-Stream Feature Fusion for\n  Category-Level Pose Estimation", "abstract": "  Category-level object pose estimation, aiming to predict the 6D pose and 3D\nsize of objects from known categories, typically struggles with large\nintra-class shape variation. Existing works utilizing mean shapes often fall\nshort of capturing this variation. To address this issue, we present\nSecondPose, a novel approach integrating object-specific geometric features\nwith semantic category priors from DINOv2. Leveraging the advantage of DINOv2\nin providing SE(3)-consistent semantic features, we hierarchically extract two\ntypes of SE(3)-invariant geometric features to further encapsulate\nlocal-to-global object-specific information. These geometric features are then\npoint-aligned with DINOv2 features to establish a consistent object\nrepresentation under SE(3) transformations, facilitating the mapping from\ncamera space to the pre-defined canonical space, thus further enhancing pose\nestimation. Extensive experiments on NOCS-REAL275 demonstrate that SecondPose\nachieves a 12.4% leap forward over the state-of-the-art. Moreover, on a more\ncomplex dataset HouseCat6D which provides photometrically challenging objects,\nSecondPose still surpasses other competitors by a large margin.\n", "date": "2023-11-18", "categories": []}
{"id": "2403.08019", "title": "MRC-Net: 6-DoF Pose Estimation with MultiScale Residual Correlation", "abstract": "  We propose a single-shot approach to determining 6-DoF pose of an object with\navailable 3D computer-aided design (CAD) model from a single RGB image. Our\nmethod, dubbed MRC-Net, comprises two stages. The first performs pose\nclassification and renders the 3D object in the classified pose. The second\nstage performs regression to predict fine-grained residual pose within class.\nConnecting the two stages is a novel multi-scale residual correlation (MRC)\nlayer that captures high-and-low level correspondences between the input image\nand rendering from first stage. MRC-Net employs a Siamese network with shared\nweights between both stages to learn embeddings for input and rendered images.\nTo mitigate ambiguity when predicting discrete pose class labels on symmetric\nobjects, we use soft probabilistic labels to define pose class in the first\nstage. We demonstrate state-of-the-art accuracy, outperforming all competing\nRGB-based methods on four challenging BOP benchmark datasets: T-LESS, LM-O,\nYCB-V, and ITODD. Our method is non-iterative and requires no complex\npost-processing.\n", "date": "2024-03-12", "categories": []}
{"id": "2403.13434", "title": "Advancing 6D Pose Estimation in Augmented Reality -- Overcoming\n  Projection Ambiguity with Uncontrolled Imagery", "abstract": "  This study addresses the challenge of accurate 6D pose estimation in\nAugmented Reality (AR), a critical component for seamlessly integrating virtual\nobjects into real-world environments. Our research primarily addresses the\ndifficulty of estimating 6D poses from uncontrolled RGB images, a common\nscenario in AR applications, which lacks metadata such as focal length. We\npropose a novel approach that strategically decomposes the estimation of z-axis\ntranslation and focal length, leveraging the neural-render and compare strategy\ninherent in the FocalPose architecture. This methodology not only streamlines\nthe 6D pose estimation process but also significantly enhances the accuracy of\n3D object overlaying in AR settings. Our experimental results demonstrate a\nmarked improvement in 6D pose estimation accuracy, with promising applications\nin manufacturing and robotics. Here, the precise overlay of AR visualizations\nand the advancement of robotic vision systems stand to benefit substantially\nfrom our findings.\n", "date": "2024-03-20", "categories": []}
{"id": "2403.12396", "title": "OV9D: Open-Vocabulary Category-Level 9D Object Pose and Size Estimation", "abstract": "  This paper studies a new open-set problem, the open-vocabulary category-level\nobject pose and size estimation. Given human text descriptions of arbitrary\nnovel object categories, the robot agent seeks to predict the position,\norientation, and size of the target object in the observed scene image. To\nenable such generalizability, we first introduce OO3D-9D, a large-scale\nphotorealistic dataset for this task. Derived from OmniObject3D, OO3D-9D is the\nlargest and most diverse dataset in the field of category-level object pose and\nsize estimation. It includes additional annotations for the symmetry axis of\neach category, which help resolve symmetric ambiguity. Apart from the\nlarge-scale dataset, we find another key to enabling such generalizability is\nleveraging the strong prior knowledge in pre-trained visual-language foundation\nmodels. We then propose a framework built on pre-trained DinoV2 and\ntext-to-image stable diffusion models to infer the normalized object coordinate\nspace (NOCS) maps of the target instances. This framework fully leverages the\nvisual semantic prior from DinoV2 and the aligned visual and language knowledge\nwithin the text-to-image diffusion model, which enables generalization to\nvarious text descriptions of novel categories. Comprehensive quantitative and\nqualitative experiments demonstrate that the proposed open-vocabulary method,\ntrained on our large-scale synthesized data, significantly outperforms the\nbaseline and can effectively generalize to real-world images of unseen\ncategories. The project page is at https://ov9d.github.io.\n", "date": "2024-03-19", "categories": []}
{"id": "2403.12682", "title": "IFFNeRF: Initialisation Free and Fast 6DoF pose estimation from a single\n  image and a NeRF model", "abstract": "  We introduce IFFNeRF to estimate the six degrees-of-freedom (6DoF) camera\npose of a given image, building on the Neural Radiance Fields (NeRF)\nformulation. IFFNeRF is specifically designed to operate in real-time and\neliminates the need for an initial pose guess that is proximate to the sought\nsolution. IFFNeRF utilizes the Metropolis-Hasting algorithm to sample surface\npoints from within the NeRF model. From these sampled points, we cast rays and\ndeduce the color for each ray through pixel-level view synthesis. The camera\npose can then be estimated as the solution to a Least Squares problem by\nselecting correspondences between the query image and the resulting bundle. We\nfacilitate this process through a learned attention mechanism, bridging the\nquery image embedding with the embedding of parameterized rays, thereby\nmatching rays pertinent to the image. Through synthetic and real evaluation\nsettings, we show that our method can improve the angular and translation error\naccuracy by 80.1% and 67.3%, respectively, compared to iNeRF while performing\nat 34fps on consumer hardware and not requiring the initial pose guess.\n", "date": "2024-03-19", "categories": []}
{"id": "2403.12728", "title": "Diffusion-Driven Self-Supervised Learning for Shape Reconstruction and\n  Pose Estimation", "abstract": "  Fully-supervised category-level pose estimation aims to determine the 6-DoF\nposes of unseen instances from known categories, requiring expensive mannual\nlabeling costs. Recently, various self-supervised category-level pose\nestimation methods have been proposed to reduce the requirement of the\nannotated datasets. However, most methods rely on synthetic data or 3D CAD\nmodel for self-supervised training, and they are typically limited to\naddressing single-object pose problems without considering multi-objective\ntasks or shape reconstruction. To overcome these challenges and limitations, we\nintroduce a diffusion-driven self-supervised network for multi-object shape\nreconstruction and categorical pose estimation, only leveraging the shape\npriors. Specifically, to capture the SE(3)-equivariant pose features and 3D\nscale-invariant shape information, we present a Prior-Aware Pyramid 3D Point\nTransformer in our network. This module adopts a point convolutional layer with\nradial-kernels for pose-aware learning and a 3D scale-invariant graph\nconvolution layer for object-level shape representation, respectively.\nFurthermore, we introduce a pretrain-to-refine self-supervised training\nparadigm to train our network. It enables proposed network to capture the\nassociations between shape priors and observations, addressing the challenge of\nintra-class shape variations by utilising the diffusion mechanism. Extensive\nexperiments conducted on four public datasets and a self-built dataset\ndemonstrate that our method significantly outperforms state-of-the-art\nself-supervised category-level baselines and even surpasses some\nfully-supervised instance-level and category-level methods.\n", "date": "2024-03-19", "categories": []}
{"id": "2403.10773", "title": "DPPE: Dense Pose Estimation in a Plenoxels Environment using Gradient\n  Approximation", "abstract": "  We present DPPE, a dense pose estimation algorithm that functions over a\nPlenoxels environment. Recent advances in neural radiance field techniques have\nshown that it is a powerful tool for environment representation. More recent\nneural rendering algorithms have significantly improved both training duration\nand rendering speed. Plenoxels introduced a fully-differentiable radiance field\ntechnique that uses Plenoptic volume elements contained in voxels for\nrendering, offering reduced training times and better rendering accuracy, while\nalso eliminating the neural net component. In this work, we introduce a 6-DoF\nmonocular RGB-only pose estimation procedure for Plenoxels, which seeks to\nrecover the ground truth camera pose after a perturbation. We employ a\nvariation on classical template matching techniques, using stochastic gradient\ndescent to optimize the pose by minimizing errors in re-rendering. In\nparticular, we examine an approach that takes advantage of the rapid rendering\nspeed of Plenoxels to numerically approximate part of the pose gradient, using\na central differencing technique. We show that such methods are effective in\npose estimation. Finally, we perform ablations over key components of the\nproblem space, with a particular focus on image subsampling and Plenoxel grid\nresolution. Project website: https://sites.google.com/view/dppe\n", "date": "2024-03-16", "categories": []}
{"id": "2403.11510", "title": "GenFlow: Generalizable Recurrent Flow for 6D Pose Refinement of Novel\n  Objects", "abstract": "  Despite the progress of learning-based methods for 6D object pose estimation,\nthe trade-off between accuracy and scalability for novel objects still exists.\nSpecifically, previous methods for novel objects do not make good use of the\ntarget object's 3D shape information since they focus on generalization by\nprocessing the shape indirectly, making them less effective. We present\nGenFlow, an approach that enables both accuracy and generalization to novel\nobjects with the guidance of the target object's shape. Our method predicts\noptical flow between the rendered image and the observed image and refines the\n6D pose iteratively. It boosts the performance by a constraint of the 3D shape\nand the generalizable geometric knowledge learned from an end-to-end\ndifferentiable system. We further improve our model by designing a cascade\nnetwork architecture to exploit the multi-scale correlations and coarse-to-fine\nrefinement. GenFlow ranked first on the unseen object pose estimation\nbenchmarks in both the RGB and RGB-D cases. It also achieves performance\ncompetitive with existing state-of-the-art methods for the seen object pose\nestimation without any fine-tuning.\n", "date": "2024-03-18", "categories": []}
{"id": "2311.14155", "title": "GigaPose: Fast and Robust Novel Object Pose Estimation via One\n  Correspondence", "abstract": "  We present GigaPose, a fast, robust, and accurate method for CAD-based novel\nobject pose estimation in RGB images. GigaPose first leverages discriminative\n\"templates\", rendered images of the CAD models, to recover the out-of-plane\nrotation and then uses patch correspondences to estimate the four remaining\nparameters. Our approach samples templates in only a two-degrees-of-freedom\nspace instead of the usual three and matches the input image to the templates\nusing fast nearest-neighbor search in feature space, results in a speedup\nfactor of 35x compared to the state of the art. Moreover, GigaPose is\nsignificantly more robust to segmentation errors. Our extensive evaluation on\nthe seven core datasets of the BOP challenge demonstrates that it achieves\nstate-of-the-art accuracy and can be seamlessly integrated with existing\nrefinement methods. Additionally, we show the potential of GigaPose with 3D\nmodels predicted by recent work on 3D reconstruction from a single image,\nrelaxing the need for CAD models and making 6D pose object estimation much more\nconvenient. Our source code and trained models are publicly available at\nhttps://github.com/nv-nguyen/gigaPose\n", "date": "2023-11-23", "categories": []}
{"id": "2403.09317", "title": "SD-Net: Symmetric-Aware Keypoint Prediction and Domain Adaptation for 6D\n  Pose Estimation In Bin-picking Scenarios", "abstract": "  Despite the success in 6D pose estimation in bin-picking scenarios, existing\nmethods still struggle to produce accurate prediction results for symmetry\nobjects and real world scenarios. The primary bottlenecks include 1) the\nambiguity keypoints caused by object symmetries; 2) the domain gap between real\nand synthetic data. To circumvent these problem, we propose a new 6D pose\nestimation network with symmetric-aware keypoint prediction and self-training\ndomain adaptation (SD-Net). SD-Net builds on pointwise keypoint regression and\ndeep hough voting to perform reliable detection keypoint under clutter and\nocclusion. Specifically, at the keypoint prediction stage, we designe a robust\n3D keypoints selection strategy considering the symmetry class of objects and\nequivalent keypoints, which facilitate locating 3D keypoints even in highly\noccluded scenes. Additionally, we build an effective filtering algorithm on\npredicted keypoint to dynamically eliminate multiple ambiguity and outlier\nkeypoint candidates. At the domain adaptation stage, we propose the\nself-training framework using a student-teacher training scheme. To carefully\ndistinguish reliable predictions, we harnesses a tailored heuristics for 3D\ngeometry pseudo labelling based on semi-chamfer distance. On public Sil'eane\ndataset, SD-Net achieves state-of-the-art results, obtaining an average\nprecision of 96%. Testing learning and generalization abilities on public\nParametric datasets, SD-Net is 8% higher than the state-of-the-art method. The\ncode is available at https://github.com/dingthuang/SD-Net.\n", "date": "2024-03-14", "categories": []}
{"id": "2403.07437", "title": "Category-Agnostic Pose Estimation for Point Clouds", "abstract": "  The goal of object pose estimation is to visually determine the pose of a\nspecific object in the RGB-D input. Unfortunately, when faced with new\ncategories, both instance-based and category-based methods are unable to deal\nwith unseen objects of unseen categories, which is a challenge for pose\nestimation. To address this issue, this paper proposes a method to introduce\ngeometric features for pose estimation of point clouds without requiring\ncategory information. The method is based only on the patch feature of the\npoint cloud, a geometric feature with rotation invariance. After training\nwithout category information, our method achieves as good results as other\ncategory-based methods. Our method successfully achieved pose annotation of no\ncategory information instances on the CAMERA25 dataset and ModelNet40 dataset.\n", "date": "2024-03-12", "categories": []}
{"id": "2311.15707", "title": "SAM-6D: Segment Anything Model Meets Zero-Shot 6D Object Pose Estimation", "abstract": "  Zero-shot 6D object pose estimation involves the detection of novel objects\nwith their 6D poses in cluttered scenes, presenting significant challenges for\nmodel generalizability. Fortunately, the recent Segment Anything Model (SAM)\nhas showcased remarkable zero-shot transfer performance, which provides a\npromising solution to tackle this task. Motivated by this, we introduce SAM-6D,\na novel framework designed to realize the task through two steps, including\ninstance segmentation and pose estimation. Given the target objects, SAM-6D\nemploys two dedicated sub-networks, namely Instance Segmentation Model (ISM)\nand Pose Estimation Model (PEM), to perform these steps on cluttered RGB-D\nimages. ISM takes SAM as an advanced starting point to generate all possible\nobject proposals and selectively preserves valid ones through meticulously\ncrafted object matching scores in terms of semantics, appearance and geometry.\nBy treating pose estimation as a partial-to-partial point matching problem, PEM\nperforms a two-stage point matching process featuring a novel design of\nbackground tokens to construct dense 3D-3D correspondence, ultimately yielding\nthe pose estimates. Without bells and whistles, SAM-6D outperforms the existing\nmethods on the seven core datasets of the BOP Benchmark for both instance\nsegmentation and pose estimation of novel objects.\n", "date": "2023-11-27", "categories": []}
{"id": "2403.03221", "title": "FAR: Flexible, Accurate and Robust 6DoF Relative Camera Pose Estimation", "abstract": "  Estimating relative camera poses between images has been a central problem in\ncomputer vision. Methods that find correspondences and solve for the\nfundamental matrix offer high precision in most cases. Conversely, methods\npredicting pose directly using neural networks are more robust to limited\noverlap and can infer absolute translation scale, but at the expense of reduced\nprecision. We show how to combine the best of both methods; our approach yields\nresults that are both precise and robust, while also accurately inferring\ntranslation scales. At the heart of our model lies a Transformer that (1)\nlearns to balance between solved and learned pose estimations, and (2) provides\na prior to guide a solver. A comprehensive analysis supports our design choices\nand demonstrates that our method adapts flexibly to various feature extractors\nand correspondence estimators, showing state-of-the-art performance in 6DoF\npose estimation on Matterport3D, InteriorNet, StreetLearn, and Map-free\nRelocalization.\n", "date": "2024-03-05", "categories": []}
{"id": "2307.11543", "title": "KVN: Keypoints Voting Network with Differentiable RANSAC for Stereo Pose\n  Estimation", "abstract": "  Object pose estimation is a fundamental computer vision task exploited in\nseveral robotics and augmented reality applications. Many established\napproaches rely on predicting 2D-3D keypoint correspondences using RANSAC\n(Random sample consensus) and estimating the object pose using the PnP\n(Perspective-n-Point) algorithm. Being RANSAC non-differentiable,\ncorrespondences cannot be directly learned in an end-to-end fashion. In this\npaper, we address the stereo image-based object pose estimation problem by i)\nintroducing a differentiable RANSAC layer into a well-known monocular pose\nestimation network; ii) exploiting an uncertainty-driven multi-view PnP solver\nwhich can fuse information from multiple views. We evaluate our approach on a\nchallenging public stereo object pose estimation dataset and a custom-built\ndataset we call Transparent Tableware Dataset (TTD), yielding state-of-the-art\nresults against other recent approaches. Furthermore, in our ablation study, we\nshow that the differentiable RANSAC layer plays a significant role in the\naccuracy of the proposed method. We release with this paper the code of our\nmethod and the TTD dataset.\n", "date": "2023-07-21", "categories": []}
{"id": "2209.03910", "title": "PixTrack: Precise 6DoF Object Pose Tracking using NeRF Templates and\n  Feature-metric Alignment", "abstract": "  We present PixTrack, a vision based object pose tracking framework using\nnovel view synthesis and deep feature-metric alignment. We follow an SfM-based\nrelocalization paradigm where we use a Neural Radiance Field to canonically\nrepresent the tracked object. Our evaluations demonstrate that our method\nproduces highly accurate, robust, and jitter-free 6DoF pose estimates of\nobjects in both monocular RGB images and RGB-D images without the need of any\ndata annotation or trajectory smoothing. Our method is also computationally\nefficient making it easy to have multi-object tracking with no alteration to\nour algorithm through simple CPU multiprocessing. Our code is available at:\nhttps://github.com/GiantAI/pixtrack\n", "date": "2022-09-08", "categories": []}
{"id": "2402.06436", "title": "Improving 2D-3D Dense Correspondences with Diffusion Models for 6D\n  Object Pose Estimation", "abstract": "  Estimating 2D-3D correspondences between RGB images and 3D space is a\nfundamental problem in 6D object pose estimation. Recent pose estimators use\ndense correspondence maps and Point-to-Point algorithms to estimate object\nposes. The accuracy of pose estimation depends heavily on the quality of the\ndense correspondence maps and their ability to withstand occlusion, clutter,\nand challenging material properties. Currently, dense correspondence maps are\nestimated using image-to-image translation models based on GANs, Autoencoders,\nor direct regression models. However, recent advancements in image-to-image\ntranslation have led to diffusion models being the superior choice when\nevaluated on benchmarking datasets. In this study, we compare image-to-image\ntranslation networks based on GANs and diffusion models for the downstream task\nof 6D object pose estimation. Our results demonstrate that the diffusion-based\nimage-to-image translation model outperforms the GAN, revealing potential for\nfurther improvements in 6D object pose estimation models.\n", "date": "2024-02-09", "categories": []}
{"id": "2402.04953", "title": "4-Dimensional deformation part model for pose estimation using Kalman\n  filter constraints", "abstract": "  The main goal of this article is to analyze the effect on pose estimation\naccuracy when using a Kalman filter added to 4-dimensional deformation part\nmodel partial solutions. The experiments run with two data sets showing that\nthis method improves pose estimation accuracy compared with state-of-the-art\nmethods and that a Kalman filter helps to increase this accuracy.\n", "date": "2024-02-07", "categories": []}
{"id": "2402.04979", "title": "Detection and Pose Estimation of flat, Texture-less Industry Objects on\n  HoloLens using synthetic Training", "abstract": "  Current state-of-the-art 6d pose estimation is too compute intensive to be\ndeployed on edge devices, such as Microsoft HoloLens (2) or Apple iPad, both\nused for an increasing number of augmented reality applications. The quality of\nAR is greatly dependent on its capabilities to detect and overlay geometry\nwithin the scene. We propose a synthetically trained client-server-based\naugmented reality application, demonstrating state-of-the-art object pose\nestimation of metallic and texture-less industry objects on edge devices.\nSynthetic data enables training without real photographs, i.e. for\nyet-to-be-manufactured objects. Our qualitative evaluation on an AR-assisted\nsorting task, and quantitative evaluation on both renderings, as well as\nreal-world data recorded on HoloLens 2, sheds light on its real-world\napplicability.\n", "date": "2024-02-07", "categories": []}
{"id": "2211.16290", "title": "LocPoseNet: Robust Location Prior for Unseen Object Pose Estimation", "abstract": "  Object location prior is critical for the standard 6D object pose estimation\nsetting. The prior can be used to initialize the 3D object translation and\nfacilitate 3D object rotation estimation. Unfortunately, the object detectors\nthat are used for this purpose do not generalize to unseen objects. Therefore,\nexisting 6D pose estimation methods for unseen objects either assume the\nground-truth object location to be known or yield inaccurate results when it is\nunavailable. In this paper, we address this problem by developing a method,\nLocPoseNet, able to robustly learn location prior for unseen objects. Our\nmethod builds upon a template matching strategy, where we propose to distribute\nthe reference kernels and convolve them with a query to efficiently compute\nmulti-scale correlations. We then introduce a novel translation estimator,\nwhich decouples scale-aware and scale-robust features to predict different\nobject location parameters. Our method outperforms existing works by a large\nmargin on LINEMOD and GenMOP. We further construct a challenging synthetic\ndataset, which allows us to highlight the better robustness of our method to\nvarious noise sources. Our project website is at:\nhttps://sailor-z.github.io/projects/3DV2024_LocPoseNet.html.\n", "date": "2022-11-29", "categories": []}
{"id": "2402.02800", "title": "Extreme Two-View Geometry From Object Poses with Diffusion Models", "abstract": "  Human has an incredible ability to effortlessly perceive the viewpoint\ndifference between two images containing the same object, even when the\nviewpoint change is astonishingly vast with no co-visible regions in the\nimages. This remarkable skill, however, has proven to be a challenge for\nexisting camera pose estimation methods, which often fail when faced with large\nviewpoint differences due to the lack of overlapping local features for\nmatching. In this paper, we aim to effectively harness the power of object\npriors to accurately determine two-view geometry in the face of extreme\nviewpoint changes. In our method, we first mathematically transform the\nrelative camera pose estimation problem to an object pose estimation problem.\nThen, to estimate the object pose, we utilize the object priors learned from a\ndiffusion model Zero123 to synthesize novel-view images of the object. The\nnovel-view images are matched to determine the object pose and thus the\ntwo-view camera pose. In experiments, our method has demonstrated extraordinary\nrobustness and resilience to large viewpoint changes, consistently estimating\ntwo-view poses with exceptional generalization ability across both synthetic\nand real-world datasets. Code will be available at\nhttps://github.com/scy639/Extreme-Two-View-Geometry-From-Object-Poses-with-Diffusion-Models.\n", "date": "2024-02-05", "categories": []}
{"id": "2401.16284", "title": "Leveraging Positional Encoding for Robust Multi-Reference-Based Object\n  6D Pose Estimation", "abstract": "  Accurately estimating the pose of an object is a crucial task in computer\nvision and robotics. There are two main deep learning approaches for this:\ngeometric representation regression and iterative refinement. However, these\nmethods have some limitations that reduce their effectiveness. In this paper,\nwe analyze these limitations and propose new strategies to overcome them. To\ntackle the issue of blurry geometric representation, we use positional encoding\nwith high-frequency components for the object's 3D coordinates. To address the\nlocal minimum problem in refinement methods, we introduce a normalized image\nplane-based multi-reference refinement strategy that's independent of intrinsic\nmatrix constraints. Lastly, we utilize adaptive instance normalization and a\nsimple occlusion augmentation method to help our model concentrate on the\ntarget object. Our experiments on Linemod, Linemod-Occlusion, and YCB-Video\ndatasets demonstrate that our approach outperforms existing methods. We will\nsoon release the code.\n", "date": "2024-01-29", "categories": []}
{"id": "2401.10848", "title": "Source-Free and Image-Only Unsupervised Domain Adaptation for Category\n  Level Object Pose Estimation", "abstract": "  We consider the problem of source-free unsupervised category-level pose\nestimation from only RGB images to a target domain without any access to source\ndomain data or 3D annotations during adaptation. Collecting and annotating\nreal-world 3D data and corresponding images is laborious, expensive, yet\nunavoidable process, since even 3D pose domain adaptation methods require 3D\ndata in the target domain. We introduce 3DUDA, a method capable of adapting to\na nuisance-ridden target domain without 3D or depth data. Our key insight stems\nfrom the observation that specific object subparts remain stable across\nout-of-domain (OOD) scenarios, enabling strategic utilization of these\ninvariant subcomponents for effective model updates. We represent object\ncategories as simple cuboid meshes, and harness a generative model of neural\nfeature activations modeled at each mesh vertex learnt using differential\nrendering. We focus on individual locally robust mesh vertex features and\niteratively update them based on their proximity to corresponding features in\nthe target domain even when the global pose is not correct. Our model is then\ntrained in an EM fashion, alternating between updating the vertex features and\nthe feature extractor. We show that our method simulates fine-tuning on a\nglobal pseudo-labeled dataset under mild assumptions, which converges to the\ntarget domain asymptotically. Through extensive empirical validation, including\na complex extreme UDA setup which combines real nuisances, synthetic noise, and\nocclusion, we demonstrate the potency of our simple approach in addressing the\ndomain shift challenge and significantly improving pose estimation accuracy.\n", "date": "2024-01-19", "categories": []}
{"id": "2401.04377", "title": "Towards Real-World Aerial Vision Guidance with Categorical 6D Pose\n  Tracker", "abstract": "  Tracking the object 6-DoF pose is crucial for various downstream robot tasks\nand real-world applications. In this paper, we investigate the real-world robot\ntask of aerial vision guidance for aerial robotics manipulation, utilizing\ncategory-level 6-DoF pose tracking. Aerial conditions inevitably introduce\nspecial challenges, such as rapid viewpoint changes in pitch and roll and\ninter-frame differences. To support these challenges in task, we firstly\nintroduce a robust category-level 6-DoF pose tracker (Robust6DoF). This tracker\nleverages shape and temporal prior knowledge to explore optimal inter-frame\nkeypoint pairs, generated under a priori structural adaptive supervision in a\ncoarse-to-fine manner. Notably, our Robust6DoF employs a Spatial-Temporal\nAugmentation module to deal with the problems of the inter-frame differences\nand intra-class shape variations through both temporal dynamic filtering and\nshape-similarity filtering. We further present a Pose-Aware Discrete Servo\nstrategy (PAD-Servo), serving as a decoupling approach to implement the final\naerial vision guidance task. It contains two servo action policies to better\naccommodate the structural properties of aerial robotics manipulation.\nExhaustive experiments on four well-known public benchmarks demonstrate the\nsuperiority of our Robust6DoF. Real-world tests directly verify that our\nRobust6DoF along with PAD-Servo can be readily used in real-world aerial\nrobotic applications.\n", "date": "2024-01-09", "categories": []}
{"id": "2401.02357", "title": "Fit-NGP: Fitting Object Models to Neural Graphics Primitives", "abstract": "  Accurate 3D object pose estimation is key to enabling many robotic\napplications that involve challenging object interactions. In this work, we\nshow that the density field created by a state-of-the-art efficient radiance\nfield reconstruction method is suitable for highly accurate and robust pose\nestimation for objects with known 3D models, even when they are very small and\nwith challenging reflective surfaces. We present a fully automatic object pose\nestimation system based on a robot arm with a single wrist-mounted camera,\nwhich can scan a scene from scratch, detect and estimate the 6-Degrees of\nFreedom (DoF) poses of multiple objects within a couple of minutes of\noperation. Small objects such as bolts and nuts are estimated with accuracy on\norder of 1mm.\n", "date": "2024-01-04", "categories": []}
{"id": "2303.05308", "title": "SpyroPose: SE(3) Pyramids for Object Pose Distribution Estimation", "abstract": "  Object pose estimation is a core computer vision problem and often an\nessential component in robotics. Pose estimation is usually approached by\nseeking the single best estimate of an object's pose, but this approach is\nill-suited for tasks involving visual ambiguity. In such cases it is desirable\nto estimate the uncertainty as a pose distribution to allow downstream tasks to\nmake informed decisions. Pose distributions can have arbitrary complexity which\nmotivates estimating unparameterized distributions, however, until now they\nhave only been used for orientation estimation on SO(3) due to the difficulty\nin training on and normalizing over SE(3). We propose a novel method for pose\ndistribution estimation on SE(3). We use a hierarchical grid, a pyramid, which\nenables efficient importance sampling during training and sparse evaluation of\nthe pyramid at inference, allowing real time 6D pose distribution estimation.\nOur method outperforms state-of-the-art methods on SO(3), and to the best of\nour knowledge, we provide the first quantitative results on pose distribution\nestimation on SE(3). Code will be available at spyropose.github.io\n", "date": "2023-03-09", "categories": []}
{"id": "2306.10531", "title": "GenPose: Generative Category-level Object Pose Estimation via Diffusion\n  Models", "abstract": "  Object pose estimation plays a vital role in embodied AI and computer vision,\nenabling intelligent agents to comprehend and interact with their surroundings.\nDespite the practicality of category-level pose estimation, current approaches\nencounter challenges with partially observed point clouds, known as the\nmultihypothesis issue. In this study, we propose a novel solution by reframing\ncategorylevel object pose estimation as conditional generative modeling,\ndeparting from traditional point-to-point regression. Leveraging score-based\ndiffusion models, we estimate object poses by sampling candidates from the\ndiffusion model and aggregating them through a two-step process: filtering out\noutliers via likelihood estimation and subsequently mean-pooling the remaining\ncandidates. To avoid the costly integration process when estimating the\nlikelihood, we introduce an alternative method that trains an energy-based\nmodel from the original score-based model, enabling end-to-end likelihood\nestimation. Our approach achieves state-of-the-art performance on the REAL275\ndataset, surpassing 50% and 60% on strict 5d2cm and 5d5cm metrics,\nrespectively. Furthermore, our method demonstrates strong generalizability to\nnovel categories sharing similar symmetric properties without fine-tuning and\ncan readily adapt to object pose tracking tasks, yielding comparable results to\nthe current state-of-the-art baselines.\n", "date": "2023-06-18", "categories": []}
{"id": "2303.12787", "title": "EPro-PnP: Generalized End-to-End Probabilistic Perspective-n-Points for\n  Monocular Object Pose Estimation", "abstract": "  Locating 3D objects from a single RGB image via Perspective-n-Point (PnP) is\na long-standing problem in computer vision. Driven by end-to-end deep learning,\nrecent studies suggest interpreting PnP as a differentiable layer, allowing for\npartial learning of 2D-3D point correspondences by backpropagating the\ngradients of pose loss. Yet, learning the entire correspondences from scratch\nis highly challenging, particularly for ambiguous pose solutions, where the\nglobally optimal pose is theoretically non-differentiable w.r.t. the points. In\nthis paper, we propose the EPro-PnP, a probabilistic PnP layer for general\nend-to-end pose estimation, which outputs a distribution of pose with\ndifferentiable probability density on the SE(3) manifold. The 2D-3D coordinates\nand corresponding weights are treated as intermediate variables learned by\nminimizing the KL divergence between the predicted and target pose\ndistribution. The underlying principle generalizes previous approaches, and\nresembles the attention mechanism. EPro-PnP can enhance existing correspondence\nnetworks, closing the gap between PnP-based method and the task-specific\nleaders on the LineMOD 6DoF pose estimation benchmark. Furthermore, EPro-PnP\nhelps to explore new possibilities of network design, as we demonstrate a novel\ndeformable correspondence network with the state-of-the-art pose accuracy on\nthe nuScenes 3D object detection benchmark. Our code is available at\nhttps://github.com/tjiiv-cprg/EPro-PnP-v2.\n", "date": "2023-03-22", "categories": []}
{"id": "2312.08268", "title": "Efficient Multi-Object Pose Estimation using Multi-Resolution Deformable\n  Attention and Query Aggregation", "abstract": "  Object pose estimation is a long-standing problem in computer vision.\nRecently, attention-based vision transformer models have achieved\nstate-of-the-art results in many computer vision applications. Exploiting the\npermutation-invariant nature of the attention mechanism, a family of vision\ntransformer models formulate multi-object pose estimation as a set prediction\nproblem. However, existing vision transformer models for multi-object pose\nestimation rely exclusively on the attention mechanism. Convolutional neural\nnetworks, on the other hand, hard-wire various inductive biases into their\narchitecture. In this paper, we investigate incorporating inductive biases in\nvision transformer models for multi-object pose estimation, which facilitates\nlearning long-range dependencies while circumventing the costly global\nattention. In particular, we use multi-resolution deformable attention, where\nthe attention operation is performed only between a few deformed reference\npoints. Furthermore, we propose a query aggregation mechanism that enables\nincreasing the number of object queries without increasing the computational\ncomplexity. We evaluate the proposed model on the challenging YCB-Video dataset\nand report state-of-the-art results.\n", "date": "2023-12-13", "categories": []}
{"id": "2312.01105", "title": "S2P3: Self-Supervised Polarimetric Pose Prediction", "abstract": "  This paper proposes the first self-supervised 6D object pose prediction from\nmultimodal RGB+polarimetric images. The novel training paradigm comprises 1) a\nphysical model to extract geometric information of polarized light, 2) a\nteacher-student knowledge distillation scheme and 3) a self-supervised loss\nformulation through differentiable rendering and an invertible physical\nconstraint. Both networks leverage the physical properties of polarized light\nto learn robust geometric representations by encoding shape priors and\npolarization characteristics derived from our physical model. Geometric\npseudo-labels from the teacher support the student network without the need for\nannotated real data. Dense appearance and geometric information of objects are\nobtained through a differentiable renderer with the predicted pose for\nself-supervised direct coupling. The student network additionally features our\nproposed invertible formulation of the physical shape priors that enables\nend-to-end self-supervised training through physical constraints of derived\npolarization characteristics compared against polarimetric input images. We\nspecifically focus on photometrically challenging objects with texture-less or\nreflective surfaces and transparent materials for which the most prominent\nperformance gain is reported.\n", "date": "2023-12-02", "categories": []}
{"id": "2212.10428", "title": "HouseCat6D -- A Large-Scale Multi-Modal Category Level 6D Object\n  Perception Dataset with Household Objects in Realistic Scenarios", "abstract": "  Estimating 6D object poses is a major challenge in 3D computer vision.\nBuilding on successful instance-level approaches, research is shifting towards\ncategory-level pose estimation for practical applications. Current\ncategory-level datasets, however, fall short in annotation quality and pose\nvariety. Addressing this, we introduce HouseCat6D, a new category-level 6D pose\ndataset. It features 1) multi-modality with Polarimetric RGB and Depth\n(RGBD+P), 2) encompasses 194 diverse objects across 10 household categories,\nincluding two photometrically challenging ones, and 3) provides high-quality\npose annotations with an error range of only 1.35 mm to 1.74 mm. The dataset\nalso includes 4) 41 large-scale scenes with comprehensive viewpoint and\nocclusion coverage, 5) a checkerboard-free environment, and 6) dense 6D\nparallel-jaw robotic grasp annotations. Additionally, we present benchmark\nresults for leading category-level pose estimation networks.\n", "date": "2022-12-20", "categories": []}
{"id": "2311.17504", "title": "PViT-6D: Overclocking Vision Transformers for 6D Pose Estimation with\n  Confidence-Level Prediction and Pose Tokens", "abstract": "  In the current state of 6D pose estimation, top-performing techniques depend\non complex intermediate correspondences, specialized architectures, and\nnon-end-to-end algorithms. In contrast, our research reframes the problem as a\nstraightforward regression task by exploring the capabilities of Vision\nTransformers for direct 6D pose estimation through a tailored use of\nclassification tokens. We also introduce a simple method for determining pose\nconfidence, which can be readily integrated into most 6D pose estimation\nframeworks. This involves modifying the transformer architecture by decreasing\nthe number of query elements based on the network's assessment of the scene\ncomplexity. Our method that we call Pose Vision Transformer or PViT-6D provides\nthe benefits of simple implementation and being end-to-end learnable while\noutperforming current state-of-the-art methods by +0.3% ADD(-S) on\nLinemod-Occlusion and +2.7% ADD(-S) on the YCB-V dataset. Moreover, our method\nenhances both the model's interpretability and the reliability of its\nperformance during inference.\n", "date": "2023-11-29", "categories": []}
{"id": "2311.13777", "title": "GS-Pose: Category-Level Object Pose Estimation via Geometric and\n  Semantic Correspondence", "abstract": "  Category-level pose estimation is a challenging task with many potential\napplications in computer vision and robotics. Recently, deep-learning-based\napproaches have made great progress, but are typically hindered by the need for\nlarge datasets of either pose-labelled real images or carefully tuned\nphotorealistic simulators. This can be avoided by using only geometry inputs\nsuch as depth images to reduce the domain-gap but these approaches suffer from\na lack of semantic information, which can be vital in the pose estimation\nproblem. To resolve this conflict, we propose to utilize both geometric and\nsemantic features obtained from a pre-trained foundation model.Our approach\nprojects 2D features from this foundation model into 3D for a single object\nmodel per category, and then performs matching against this for new single view\nobservations of unseen object instances with a trained matching network. This\nrequires significantly less data to train than prior methods since the semantic\nfeatures are robust to object texture and appearance. We demonstrate this with\na rich evaluation, showing improved performance over prior methods with a\nfraction of the data required.\n", "date": "2023-11-23", "categories": []}
{"id": "2305.10808", "title": "Manifold-Aware Self-Training for Unsupervised Domain Adaptation on\n  Regressing 6D Object Pose", "abstract": "  Domain gap between synthetic and real data in visual regression (e.g. 6D pose\nestimation) is bridged in this paper via global feature alignment and local\nrefinement on the coarse classification of discretized anchor classes in target\nspace, which imposes a piece-wise target manifold regularization into\ndomain-invariant representation learning. Specifically, our method incorporates\nan explicit self-supervised manifold regularization, revealing consistent\ncumulative target dependency across domains, to a self-training scheme (e.g.\nthe popular Self-Paced Self-Training) to encourage more discriminative\ntransferable representations of regression tasks. Moreover, learning unified\nimplicit neural functions to estimate relative direction and distance of\ntargets to their nearest class bins aims to refine target classification\npredictions, which can gain robust performance against inconsistent feature\nscaling sensitive to UDA regressors. Experiment results on three public\nbenchmarks of the challenging 6D pose estimation task can verify the\neffectiveness of our method, consistently achieving superior performance to the\nstate-of-the-art for UDA on 6D pose estimation.\n", "date": "2023-05-18", "categories": []}
{"id": "2311.09269", "title": "NormNet: Scale Normalization for 6D Pose Estimation in Stacked Scenarios", "abstract": "  Existing Object Pose Estimation (OPE) methods for stacked scenarios are not\nrobust to changes in object scale. This paper proposes a new 6DoF OPE network\n(NormNet) for different scale objects in stacked scenarios. Specifically, each\nobject's scale is first learned with point-wise regression. Then, all objects\nin the stacked scenario are normalized into the same scale through semantic\nsegmentation and affine transformation. Finally, they are fed into a shared\npose estimator to recover their 6D poses. In addition, we introduce a new\nSim-to-Real transfer pipeline, combining style transfer and domain\nrandomization. This improves the NormNet's performance on real data even if we\nonly train it on synthetic data. Extensive experiments demonstrate that the\nproposed method achieves state-of-the-art performance on public benchmarks and\nthe MultiScale dataset we constructed. The real-world experiments show that our\nmethod can robustly estimate the 6D pose of objects at different scales.\n", "date": "2023-11-15", "categories": []}
{"id": "2311.08438", "title": "LocaliseBot: Multi-view 3D object localisation with differentiable\n  rendering for robot grasping", "abstract": "  Robot grasp typically follows five stages: object detection, object\nlocalisation, object pose estimation, grasp pose estimation, and grasp\nplanning. We focus on object pose estimation. Our approach relies on three\npieces of information: multiple views of the object, the camera's extrinsic\nparameters at those viewpoints, and 3D CAD models of objects. The first step\ninvolves a standard deep learning backbone (FCN ResNet) to estimate the object\nlabel, semantic segmentation, and a coarse estimate of the object pose with\nrespect to the camera. Our novelty is using a refinement module that starts\nfrom the coarse pose estimate and refines it by optimisation through\ndifferentiable rendering. This is a purely vision-based approach that avoids\nthe need for other information such as point cloud or depth images. We evaluate\nour object pose estimation approach on the ShapeNet dataset and show\nimprovements over the state of the art. We also show that the estimated object\npose results in 99.65% grasp accuracy with the ground truth grasp candidates on\nthe Object Clutter Indoor Dataset (OCID) Grasp dataset, as computed using\nstandard practice.\n", "date": "2023-11-14", "categories": []}
{"id": "2307.05016", "title": "TRansPose: Large-Scale Multispectral Dataset for Transparent Object", "abstract": "  Transparent objects are encountered frequently in our daily lives, yet\nrecognizing them poses challenges for conventional vision sensors due to their\nunique material properties, not being well perceived from RGB or depth cameras.\nOvercoming this limitation, thermal infrared cameras have emerged as a\nsolution, offering improved visibility and shape information for transparent\nobjects. In this paper, we present TRansPose, the first large-scale\nmultispectral dataset that combines stereo RGB-D, thermal infrared (TIR)\nimages, and object poses to promote transparent object research. The dataset\nincludes 99 transparent objects, encompassing 43 household items, 27 recyclable\ntrashes, 29 chemical laboratory equivalents, and 12 non-transparent objects. It\ncomprises a vast collection of 333,819 images and 4,000,056 annotations,\nproviding instance-level segmentation masks, ground-truth poses, and completed\ndepth information. The data was acquired using a FLIR A65 thermal infrared\n(TIR) camera, two Intel RealSense L515 RGB-D cameras, and a Franka Emika Panda\nrobot manipulator. Spanning 87 sequences, TRansPose covers various challenging\nreal-life scenarios, including objects filled with water, diverse lighting\nconditions, heavy clutter, non-transparent or translucent containers, objects\nin plastic bags, and multi-stacked objects. TRansPose dataset can be accessed\nfrom the following link: https://sites.google.com/view/transpose-dataset\n", "date": "2023-07-11", "categories": []}
{"id": "2308.07827", "title": "Learning Better Keypoints for Multi-Object 6DoF Pose Estimation", "abstract": "  We address the problem of keypoint selection, and find that the performance\nof 6DoF pose estimation methods can be improved when pre-defined keypoint\nlocations are learned, rather than being heuristically selected as has been the\nstandard approach. We found that accuracy and efficiency can be improved by\ntraining a graph network to select a set of disperse keypoints with similarly\ndistributed votes. These votes, learned by a regression network to accumulate\nevidence for the keypoint locations, can be regressed more accurately compared\nto previous heuristic keypoint algorithms. The proposed KeyGNet, supervised by\na combined loss measuring both Wasserstein distance and dispersion, learns the\ncolor and geometry features of the target objects to estimate optimal keypoint\nlocations. Experiments demonstrate the keypoints selected by KeyGNet improved\nthe accuracy for all evaluation metrics of all seven datasets tested, for three\nkeypoint voting methods. The challenging Occlusion LINEMOD dataset notably\nimproved ADD(S) by +16.4% on PVN3D, and all core BOP datasets showed an AR\nimprovement for all objects, of between +1% and +21.5%. There was also a\nnotable increase in performance when transitioning from single object to\nmultiple object training using KeyGNet keypoints, essentially eliminating the\nSISO-MIMO gap for Occlusion LINEMOD.\n", "date": "2023-08-15", "categories": []}
{"id": "2310.17359", "title": "SE(3) Diffusion Model-based Point Cloud Registration for Robust 6D\n  Object Pose Estimation", "abstract": "  In this paper, we introduce an SE(3) diffusion model-based point cloud\nregistration framework for 6D object pose estimation in real-world scenarios.\nOur approach formulates the 3D registration task as a denoising diffusion\nprocess, which progressively refines the pose of the source point cloud to\nobtain a precise alignment with the model point cloud. Training our framework\ninvolves two operations: An SE(3) diffusion process and an SE(3) reverse\nprocess. The SE(3) diffusion process gradually perturbs the optimal rigid\ntransformation of a pair of point clouds by continuously injecting noise\n(perturbation transformation). By contrast, the SE(3) reverse process focuses\non learning a denoising network that refines the noisy transformation\nstep-by-step, bringing it closer to the optimal transformation for accurate\npose estimation. Unlike standard diffusion models used in linear Euclidean\nspaces, our diffusion model operates on the SE(3) manifold. This requires\nexploiting the linear Lie algebra $\\mathfrak{se}(3)$ associated with SE(3) to\nconstrain the transformation transitions during the diffusion and reverse\nprocesses. Additionally, to effectively train our denoising network, we derive\na registration-specific variational lower bound as the optimization objective\nfor model learning. Furthermore, we show that our denoising network can be\nconstructed with a surrogate registration model, making our approach applicable\nto different deep registration networks. Extensive experiments demonstrate that\nour diffusion registration framework presents outstanding pose estimation\nperformance on the real-world TUD-L, LINEMOD, and Occluded-LINEMOD datasets.\n", "date": "2023-10-26", "categories": []}
{"id": "2310.16618", "title": "Real-time 6-DoF Pose Estimation by an Event-based Camera using Active\n  LED Markers", "abstract": "  Real-time applications for autonomous operations depend largely on fast and\nrobust vision-based localization systems. Since image processing tasks require\nprocessing large amounts of data, the computational resources often limit the\nperformance of other processes. To overcome this limitation, traditional\nmarker-based localization systems are widely used since they are easy to\nintegrate and achieve reliable accuracy. However, classical marker-based\nlocalization systems significantly depend on standard cameras with low frame\nrates, which often lack accuracy due to motion blur. In contrast, event-based\ncameras provide high temporal resolution and a high dynamic range, which can be\nutilized for fast localization tasks, even under challenging visual conditions.\nThis paper proposes a simple but effective event-based pose estimation system\nusing active LED markers (ALM) for fast and accurate pose estimation. The\nproposed algorithm is able to operate in real time with a latency below\n\\SI{0.5}{\\milli\\second} while maintaining output rates of \\SI{3}{\\kilo \\hertz}.\nExperimental results in static and dynamic scenarios are presented to\ndemonstrate the performance of the proposed approach in terms of computational\nspeed and absolute accuracy, using the OptiTrack system as the basis for\nmeasurement.\n", "date": "2023-10-25", "categories": []}
{"id": "2211.13572", "title": "Physics-Based Object 6D-Pose Estimation during Non-Prehensile\n  Manipulation", "abstract": "  We propose a method to track the 6D pose of an object over time, while the\nobject is under non-prehensile manipulation by a robot. At any given time\nduring the manipulation of the object, we assume access to the robot joint\ncontrols and an image from a camera. We use the robot joint controls to perform\na physics-based prediction of how the object might be moving. We then combine\nthis prediction with the observation coming from the camera, to estimate the\nobject pose as accurately as possible. We use a particle filtering approach to\ncombine the control information with the visual information. We compare the\nproposed method with two baselines: (i) using only an image-based pose\nestimation system at each time-step, and (ii) a particle filter which does not\nperform the computationally expensive physics predictions, but assumes the\nobject moves with constant velocity. Our results show that making physics-based\npredictions is worth the computational cost, resulting in more accurate\ntracking, and estimating object pose even when the object is not clearly\nvisible to the camera.\n", "date": "2022-11-24", "categories": []}
{"id": "2206.07162", "title": "Category-Agnostic 6D Pose Estimation with Conditional Neural Processes", "abstract": "  We present a novel meta-learning approach for 6D pose estimation on unknown\nobjects. In contrast to ``instance-level\" and ``category-level\" pose estimation\nmethods, our algorithm learns object representation in a category-agnostic way,\nwhich endows it with strong generalization capabilities across object\ncategories. Specifically, we employ a neural process-based meta-learning\napproach to train an encoder to capture texture and geometry of an object in a\nlatent representation, based on very few RGB-D images and ground-truth\nkeypoints. The latent representation is then used by a simultaneously\nmeta-trained decoder to predict the 6D pose of the object in new images.\nFurthermore, we propose a novel geometry-aware decoder for the keypoint\nprediction using a Graph Neural Network (GNN), which explicitly takes geometric\nconstraints specific to each object into consideration. To evaluate our\nalgorithm, extensive experiments are conducted on the \\linemod dataset, and on\nour new fully-annotated synthetic datasets generated from Multiple Categories\nin Multiple Scenes (MCMS). Experimental results demonstrate that our model\nperforms well on unseen objects with very different shapes and appearances.\nRemarkably, our model also shows robust performance on occluded scenes although\ntrained fully on data without occlusion. To our knowledge, this is the first\nwork exploring \\textbf{cross-category level} 6D pose estimation.\n", "date": "2022-06-14", "categories": []}
{"id": "2310.12974", "title": "FSD: Fast Self-Supervised Single RGB-D to Categorical 3D Objects", "abstract": "  In this work, we address the challenging task of 3D object recognition\nwithout the reliance on real-world 3D labeled data. Our goal is to predict the\n3D shape, size, and 6D pose of objects within a single RGB-D image, operating\nat the category level and eliminating the need for CAD models during inference.\nWhile existing self-supervised methods have made strides in this field, they\noften suffer from inefficiencies arising from non-end-to-end processing,\nreliance on separate models for different object categories, and slow surface\nextraction during the training of implicit reconstruction models; thus\nhindering both the speed and real-world applicability of the 3D recognition\nprocess. Our proposed method leverages a multi-stage training pipeline,\ndesigned to efficiently transfer synthetic performance to the real-world\ndomain. This approach is achieved through a combination of 2D and 3D supervised\nlosses during the synthetic domain training, followed by the incorporation of\n2D supervised and 3D self-supervised losses on real-world data in two\nadditional learning stages. By adopting this comprehensive strategy, our method\nsuccessfully overcomes the aforementioned limitations and outperforms existing\nself-supervised 6D pose and size estimation baselines on the NOCS test-set with\na 16.4% absolute improvement in mAP for 6D pose estimation while running in\nnear real-time at 5 Hz.\n", "date": "2023-10-19", "categories": []}
{"id": "2309.10255", "title": "RGB-based Category-level Object Pose Estimation via Decoupled Metric\n  Scale Recovery", "abstract": "  While showing promising results, recent RGB-D camera-based category-level\nobject pose estimation methods have restricted applications due to the heavy\nreliance on depth sensors. RGB-only methods provide an alternative to this\nproblem yet suffer from inherent scale ambiguity stemming from monocular\nobservations. In this paper, we propose a novel pipeline that decouples the 6D\npose and size estimation to mitigate the influence of imperfect scales on rigid\ntransformations. Specifically, we leverage a pre-trained monocular estimator to\nextract local geometric information, mainly facilitating the search for inlier\n2D-3D correspondence. Meanwhile, a separate branch is designed to directly\nrecover the metric scale of the object based on category-level statistics.\nFinally, we advocate using the RANSAC-P$n$P algorithm to robustly solve for 6D\nobject pose. Extensive experiments have been conducted on both synthetic and\nreal datasets, demonstrating the superior performance of our method over\nprevious state-of-the-art RGB-based approaches, especially in terms of rotation\naccuracy. Code: https://github.com/goldoak/DMSR.\n", "date": "2023-09-19", "categories": []}
{"id": "2310.11184", "title": "Sparse Multi-Object Render-and-Compare", "abstract": "  Reconstructing 3D shape and pose of static objects from a single image is an\nessential task for various industries, including robotics, augmented reality,\nand digital content creation. This can be done by directly predicting 3D shape\nin various representations or by retrieving CAD models from a database and\npredicting their alignments. Directly predicting 3D shapes often produces\nunrealistic, overly smoothed or tessellated shapes. Retrieving CAD models\nensures realistic shapes but requires robust and accurate alignment. Learning\nto directly predict CAD model poses from image features is challenging and\ninaccurate. Works, such as ROCA, compute poses from predicted normalised object\ncoordinates which can be more accurate but are susceptible to systematic\nfailure. SPARC demonstrates that following a ''render-and-compare'' approach\nwhere a network iteratively improves upon its own predictions achieves accurate\nalignments. Nevertheless, it performs individual CAD alignment for every object\ndetected in an image. This approach is slow when applied to many objects as the\ntime complexity increases linearly with the number of objects and can not learn\ninter-object relations. Introducing a new network architecture Multi-SPARC we\nlearn to perform CAD model alignments for multiple detected objects jointly.\nCompared to other single-view methods we achieve state-of-the-art performance\non the challenging real-world dataset ScanNet. By improving the instance\nalignment accuracy from 31.8% to 40.3% we perform similar to state-of-the-art\nmulti-view methods.\n", "date": "2023-10-17", "categories": []}
{"id": "2303.11516", "title": "Linear-Covariance Loss for End-to-End Learning of 6D Pose Estimation", "abstract": "  Most modern image-based 6D object pose estimation methods learn to predict\n2D-3D correspondences, from which the pose can be obtained using a PnP solver.\nBecause of the non-differentiable nature of common PnP solvers, these methods\nare supervised via the individual correspondences. To address this, several\nmethods have designed differentiable PnP strategies, thus imposing supervision\non the pose obtained after the PnP step. Here, we argue that this conflicts\nwith the averaging nature of the PnP problem, leading to gradients that may\nencourage the network to degrade the accuracy of individual correspondences. To\naddress this, we derive a loss function that exploits the ground truth pose\nbefore solving the PnP problem. Specifically, we linearize the PnP solver\naround the ground-truth pose and compute the covariance of the resulting pose\ndistribution. We then define our loss based on the diagonal covariance\nelements, which entails considering the final pose estimate yet not suffering\nfrom the PnP averaging issue. Our experiments show that our loss consistently\nimproves the pose estimation accuracy for both dense and sparse correspondence\nbased methods, achieving state-of-the-art results on both Linemod-Occluded and\nYCB-Video.\n", "date": "2023-03-21", "categories": []}
{"id": "2310.03534", "title": "3D-Aware Hypothesis & Verification for Generalizable Relative Object\n  Pose Estimation", "abstract": "  Prior methods that tackle the problem of generalizable object pose estimation\nhighly rely on having dense views of the unseen object. By contrast, we address\nthe scenario where only a single reference view of the object is available. Our\ngoal then is to estimate the relative object pose between this reference view\nand a query image that depicts the object in a different pose. In this\nscenario, robust generalization is imperative due to the presence of unseen\nobjects during testing and the large-scale object pose variation between the\nreference and the query. To this end, we present a new\nhypothesis-and-verification framework, in which we generate and evaluate\nmultiple pose hypotheses, ultimately selecting the most reliable one as the\nrelative object pose. To measure reliability, we introduce a 3D-aware\nverification that explicitly applies 3D transformations to the 3D object\nrepresentations learned from the two input images. Our comprehensive\nexperiments on the Objaverse, LINEMOD, and CO3D datasets evidence the superior\naccuracy of our approach in relative pose estimation and its robustness in\nlarge-scale pose variations, when dealing with unseen objects.\n", "date": "2023-10-05", "categories": []}
{"id": "2307.15514", "title": "Revisiting Fully Convolutional Geometric Features for Object 6D Pose\n  Estimation", "abstract": "  Recent works on 6D object pose estimation focus on learning keypoint\ncorrespondences between images and object models, and then determine the object\npose through RANSAC-based algorithms or by directly regressing the pose with\nend-to-end optimisations. We argue that learning point-level discriminative\nfeatures is overlooked in the literature. To this end, we revisit Fully\nConvolutional Geometric Features (FCGF) and tailor it for object 6D pose\nestimation to achieve state-of-the-art performance. FCGF employs sparse\nconvolutions and learns point-level features using a fully-convolutional\nnetwork by optimising a hardest contrastive loss. We can outperform recent\ncompetitors on popular benchmarks by adopting key modifications to the loss and\nto the input data representations, by carefully tuning the training strategies,\nand by employing data augmentations suitable for the underlying problem. We\ncarry out a thorough ablation to study the contribution of each modification.\nThe code is available at https://github.com/jcorsetti/FCGF6D.\n", "date": "2023-07-28", "categories": []}
{"id": "2310.01897", "title": "MFOS: Model-Free & One-Shot Object Pose Estimation", "abstract": "  Existing learning-based methods for object pose estimation in RGB images are\nmostly model-specific or category based. They lack the capability to generalize\nto new object categories at test time, hence severely hindering their\npracticability and scalability. Notably, recent attempts have been made to\nsolve this issue, but they still require accurate 3D data of the object surface\nat both train and test time. In this paper, we introduce a novel approach that\ncan estimate in a single forward pass the pose of objects never seen during\ntraining, given minimum input. In contrast to existing state-of-the-art\napproaches, which rely on task-specific modules, our proposed model is entirely\nbased on a transformer architecture, which can benefit from recently proposed\n3D-geometry general pretraining. We conduct extensive experiments and report\nstate-of-the-art one-shot performance on the challenging LINEMOD benchmark.\nFinally, extensive ablations allow us to determine good practices with this\nrelatively new type of architecture in the field.\n", "date": "2023-10-03", "categories": []}
{"id": "2305.07348", "title": "A Survey on Deep Learning-Based Monocular Spacecraft Pose Estimation:\n  Current State, Limitations and Prospects", "abstract": "  Estimating the pose of an uncooperative spacecraft is an important computer\nvision problem for enabling the deployment of automatic vision-based systems in\norbit, with applications ranging from on-orbit servicing to space debris\nremoval. Following the general trend in computer vision, more and more works\nhave been focusing on leveraging Deep Learning (DL) methods to address this\nproblem. However and despite promising research-stage results, major challenges\npreventing the use of such methods in real-life missions still stand in the\nway. In particular, the deployment of such computation-intensive algorithms is\nstill under-investigated, while the performance drop when training on synthetic\nand testing on real images remains to mitigate. The primary goal of this survey\nis to describe the current DL-based methods for spacecraft pose estimation in a\ncomprehensive manner. The secondary goal is to help define the limitations\ntowards the effective deployment of DL-based spacecraft pose estimation\nsolutions for reliable autonomous vision-based applications. To this end, the\nsurvey first summarises the existing algorithms according to two approaches:\nhybrid modular pipelines and direct end-to-end regression methods. A comparison\nof algorithms is presented not only in terms of pose accuracy but also with a\nfocus on network architectures and models' sizes keeping potential deployment\nin mind. Then, current monocular spacecraft pose estimation datasets used to\ntrain and test these methods are discussed. The data generation methods:\nsimulators and testbeds, the domain gap and the performance drop between\nsynthetically generated and lab/space collected images and the potential\nsolutions are also discussed. Finally, the paper presents open research\nquestions and future directions in the field, drawing parallels with other\ncomputer vision applications.\n", "date": "2023-05-12", "categories": []}
{"id": "2310.00463", "title": "Diff-DOPE: Differentiable Deep Object Pose Estimation", "abstract": "  We introduce Diff-DOPE, a 6-DoF pose refiner that takes as input an image, a\n3D textured model of an object, and an initial pose of the object. The method\nuses differentiable rendering to update the object pose to minimize the visual\nerror between the image and the projection of the model. We show that this\nsimple, yet effective, idea is able to achieve state-of-the-art results on pose\nestimation datasets. Our approach is a departure from recent methods in which\nthe pose refiner is a deep neural network trained on a large synthetic dataset\nto map inputs to refinement steps. Rather, our use of differentiable rendering\nallows us to avoid training altogether. Our approach performs multiple gradient\ndescent optimizations in parallel with different random learning rates to avoid\nlocal minima from symmetric objects, similar appearances, or wrong step size.\nVarious modalities can be used, e.g., RGB, depth, intensity edges, and object\nsegmentation masks. We present experiments examining the effect of various\nchoices, showing that the best results are found when the RGB image is\naccompanied by an object mask and depth image to guide the optimization\nprocess.\n", "date": "2023-09-30", "categories": []}
{"id": "2309.11986", "title": "ZS6D: Zero-shot 6D Object Pose Estimation using Vision Transformers", "abstract": "  As robotic systems increasingly encounter complex and unconstrained\nreal-world scenarios, there is a demand to recognize diverse objects. The\nstate-of-the-art 6D object pose estimation methods rely on object-specific\ntraining and therefore do not generalize to unseen objects. Recent novel object\npose estimation methods are solving this issue using task-specific fine-tuned\nCNNs for deep template matching. This adaptation for pose estimation still\nrequires expensive data rendering and training procedures. MegaPose for example\nis trained on a dataset consisting of two million images showing 20,000\ndifferent objects to reach such generalization capabilities. To overcome this\nshortcoming we introduce ZS6D, for zero-shot novel object 6D pose estimation.\nVisual descriptors, extracted using pre-trained Vision Transformers (ViT), are\nused for matching rendered templates against query images of objects and for\nestablishing local correspondences. These local correspondences enable deriving\ngeometric correspondences and are used for estimating the object's 6D pose with\nRANSAC-based PnP. This approach showcases that the image descriptors extracted\nby pre-trained ViTs are well-suited to achieve a notable improvement over two\nstate-of-the-art novel object 6D pose estimation methods, without the need for\ntask-specific fine-tuning. Experiments are performed on LMO, YCBV, and TLESS.\nIn comparison to one of the two methods we improve the Average Recall on all\nthree datasets and compared to the second method we improve on two datasets.\n", "date": "2023-09-21", "categories": []}
{"id": "2206.12498", "title": "Optimal and Robust Category-level Perception: Object Pose and Shape\n  Estimation from 2D and 3D Semantic Keypoints", "abstract": "  We consider a category-level perception problem, where one is given 2D or 3D\nsensor data picturing an object of a given category (e.g., a car), and has to\nreconstruct the 3D pose and shape of the object despite intra-class variability\n(i.e., different car models have different shapes). We consider an active shape\nmodel, where -- for an object category -- we are given a library of potential\nCAD models describing objects in that category, and we adopt a standard\nformulation where pose and shape are estimated from 2D or 3D keypoints via\nnon-convex optimization. Our first contribution is to develop PACE3D* and\nPACE2D*, the first certifiably optimal solvers for pose and shape estimation\nusing 3D and 2D keypoints, respectively. Both solvers rely on the design of\ntight (i.e., exact) semidefinite relaxations. Our second contribution is to\ndevelop outlier-robust versions of both solvers, named PACE3D# and PACE2D#.\nTowards this goal, we propose ROBIN, a general graph-theoretic framework to\nprune outliers, which uses compatibility hypergraphs to model measurements'\ncompatibility. We show that in category-level perception problems these\nhypergraphs can be built from the winding orders of the keypoints (in 2D) or\ntheir convex hulls (in 3D), and many outliers can be filtered out via maximum\nhyperclique computation. The last contribution is an extensive experimental\nevaluation. Besides providing an ablation study on simulated datasets and on\nthe PASCAL3D+ dataset, we combine our solver with a deep keypoint detector, and\nshow that PACE3D# improves over the state of the art in vehicle pose estimation\nin the ApolloScape datasets, and its runtime is compatible with practical\napplications. We release our code at https://github.com/MIT-SPARK/PACE.\n", "date": "2022-06-24", "categories": []}
{"id": "2302.06821", "title": "Model-Based Underwater 6D Pose Estimation from RGB", "abstract": "  Object pose estimation underwater allows an autonomous system to perform\ntracking and intervention tasks. Nonetheless, underwater target pose estimation\nis remarkably challenging due to, among many factors, limited visibility, light\nscattering, cluttered environments, and constantly varying water conditions. An\napproach is to employ sonar or laser sensing to acquire 3D data, however, the\ndata is not clear and the sensors expensive. For this reason, the community has\nfocused on extracting pose estimates from RGB input. In this work, we propose\nan approach that leverages 2D object detection to reliably compute 6D pose\nestimates in different underwater scenarios. We test our proposal with 4\nobjects with symmetrical shapes and poor texture spanning across 33,920\nsynthetic and 10 real scenes. All objects and scenes are made available in an\nopen-source dataset that includes annotations for object detection and pose\nestimation. When benchmarking against similar end-to-end methodologies for 6D\nobject pose estimation, our pipeline provides estimates that are 8% more\naccurate. We also demonstrate the real world usability of our pose estimation\npipeline on an underwater robotic manipulator in a reaching task.\n", "date": "2023-02-14", "categories": []}
{"id": "2309.07796", "title": "For A More Comprehensive Evaluation of 6DoF Object Pose Tracking", "abstract": "  Previous evaluations on 6DoF object pose tracking have presented obvious\nlimitations along with the development of this area. In particular, the\nevaluation protocols are not unified for different methods, the widely-used\nYCBV dataset contains significant annotation error, and the error metrics also\nmay be biased. As a result, it is hard to fairly compare the methods, which has\nbecame a big obstacle for developing new algorithms. In this paper we\ncontribute a unified benchmark to address the above problems. For more accurate\nannotation of YCBV, we propose a multi-view multi-object global pose refinement\nmethod, which can jointly refine the poses of all objects and view cameras,\nresulting in sub-pixel sub-millimeter alignment errors. The limitations of\nprevious scoring methods and error metrics are analyzed, based on which we\nintroduce our improved evaluation methods. The unified benchmark takes both\nYCBV and BCOT as base datasets, which are shown to be complementary in scene\ncategories. In experiments, we validate the precision and reliability of the\nproposed global pose refinement method with a realistic semi-synthesized\ndataset particularly for YCBV, and then present the benchmark results unifying\nlearning&non-learning and RGB&RGBD methods, with some finds not discovered in\nprevious studies.\n", "date": "2023-09-14", "categories": []}
{"id": "2308.08518", "title": "Exploiting Point-Wise Attention in 6D Object Pose Estimation Based on\n  Bidirectional Prediction", "abstract": "  Traditional geometric registration based estimation methods only exploit the\nCAD model implicitly, which leads to their dependence on observation quality\nand deficiency to occlusion. To address the problem,the paper proposes a\nbidirectional correspondence prediction network with a point-wise\nattention-aware mechanism. This network not only requires the model points to\npredict the correspondence but also explicitly models the geometric\nsimilarities between observations and the model prior. Our key insight is that\nthe correlations between each model point and scene point provide essential\ninformation for learning point-pair matches. To further tackle the correlation\nnoises brought by feature distribution divergence, we design a simple but\neffective pseudo-siamese network to improve feature homogeneity. Experimental\nresults on the public datasets of LineMOD, YCB-Video, and Occ-LineMOD show that\nthe proposed method achieves better performance than other state-of-the-art\nmethods under the same evaluation criteria. Its robustness in estimating poses\nis greatly improved, especially in an environment with severe occlusions.\n", "date": "2023-08-16", "categories": []}
{"id": "2309.05662", "title": "ViHOPE: Visuotactile In-Hand Object 6D Pose Estimation with Shape\n  Completion", "abstract": "  In this letter, we introduce ViHOPE, a novel framework for estimating the 6D\npose of an in-hand object using visuotactile perception. Our key insight is\nthat the accuracy of the 6D object pose estimate can be improved by explicitly\ncompleting the shape of the object. To this end, we introduce a novel\nvisuotactile shape completion module that uses a conditional Generative\nAdversarial Network to complete the shape of an in-hand object based on\nvolumetric representation. This approach improves over prior works that\ndirectly regress visuotactile observations to a 6D pose. By explicitly\ncompleting the shape of the in-hand object and jointly optimizing the shape\ncompletion and pose estimation tasks, we improve the accuracy of the 6D object\npose estimate. We train and test our model on a synthetic dataset and compare\nit with the state-of-the-art. In the visuotactile shape completion task, we\noutperform the state-of-the-art by 265% using the Intersection of Union metric\nand achieve 88% lower Chamfer Distance. In the visuotactile pose estimation\ntask, we present results that suggest our framework reduces position and\nangular errors by 35% and 64%, respectively. Furthermore, we ablate our\nframework to confirm the gain on the 6D object pose estimate from explicitly\ncompleting the shape. Ultimately, we show that our framework produces models\nthat are robust to sim-to-real transfer on a real-world robot platform.\n", "date": "2023-09-11", "categories": []}
{"id": "2302.03744", "title": "3D Neural Embedding Likelihood: Probabilistic Inverse Graphics for\n  Robust 6D Pose Estimation", "abstract": "  The ability to perceive and understand 3D scenes is crucial for many\napplications in computer vision and robotics. Inverse graphics is an appealing\napproach to 3D scene understanding that aims to infer the 3D scene structure\nfrom 2D images. In this paper, we introduce probabilistic modeling to the\ninverse graphics framework to quantify uncertainty and achieve robustness in 6D\npose estimation tasks. Specifically, we propose 3D Neural Embedding Likelihood\n(3DNEL) as a unified probabilistic model over RGB-D images, and develop\nefficient inference procedures on 3D scene descriptions. 3DNEL effectively\ncombines learned neural embeddings from RGB with depth information to improve\nrobustness in sim-to-real 6D object pose estimation from RGB-D images.\nPerformance on the YCB-Video dataset is on par with state-of-the-art yet is\nmuch more robust in challenging regimes. In contrast to discriminative\napproaches, 3DNEL's probabilistic generative formulation jointly models\nmultiple objects in a scene, quantifies uncertainty in a principled way, and\nhandles object pose tracking under heavy occlusion. Finally, 3DNEL provides a\nprincipled framework for incorporating prior knowledge about the scene and\nobjects, which allows natural extension to additional tasks like camera pose\ntracking from video.\n", "date": "2023-02-07", "categories": []}
{"id": "2302.10602", "title": "SU-Net: Pose estimation network for non-cooperative spacecraft on-orbit", "abstract": "  Spacecraft pose estimation plays a vital role in many on-orbit space\nmissions, such as rendezvous and docking, debris removal, and on-orbit\nmaintenance. At present, space images contain widely varying lighting\nconditions, high contrast and low resolution, pose estimation of space objects\nis more challenging than that of objects on earth. In this paper, we analyzing\nthe radar image characteristics of spacecraft on-orbit, then propose a new deep\nlearning neural Network structure named Dense Residual U-shaped Network\n(DR-U-Net) to extract image features. We further introduce a novel neural\nnetwork based on DR-U-Net, namely Spacecraft U-shaped Network (SU-Net) to\nachieve end-to-end pose estimation for non-cooperative spacecraft.\nSpecifically, the SU-Net first preprocess the image of non-cooperative\nspacecraft, then transfer learning was used for pre-training. Subsequently, in\norder to solve the problem of radar image blur and low ability of spacecraft\ncontour recognition, we add residual connection and dense connection to the\nbackbone network U-Net, and we named it DR-U-Net. In this way, the feature loss\nand the complexity of the model is reduced, and the degradation of deep neural\nnetwork during training is avoided. Finally, a layer of feedforward neural\nnetwork is used for pose estimation of non-cooperative spacecraft on-orbit.\nExperiments prove that the proposed method does not rely on the hand-made\nobject specific features, and the model has robust robustness, and the\ncalculation accuracy outperforms the state-of-the-art pose estimation methods.\nThe absolute error is 0.1557 to 0.4491 , the mean error is about 0.302 , and\nthe standard deviation is about 0.065 .\n", "date": "2023-02-21", "categories": []}
{"id": "2210.13540", "title": "Video based Object 6D Pose Estimation using Transformers", "abstract": "  We introduce a Transformer based 6D Object Pose Estimation framework\nVideoPose, comprising an end-to-end attention based modelling architecture,\nthat attends to previous frames in order to estimate accurate 6D Object Poses\nin videos. Our approach leverages the temporal information from a video\nsequence for pose refinement, along with being computationally efficient and\nrobust. Compared to existing methods, our architecture is able to capture and\nreason from long-range dependencies efficiently, thus iteratively refining over\nvideo sequences. Experimental evaluation on the YCB-Video dataset shows that\nour approach is on par with the state-of-the-art Transformer methods, and\nperforms significantly better relative to CNN based approaches. Further, with a\nspeed of 33 fps, it is also more efficient and therefore applicable to a\nvariety of applications that require real-time object pose estimation. Training\ncode and pretrained models are available at\nhttps://github.com/ApoorvaBeedu/VideoPose\n", "date": "2022-10-24", "categories": []}
{"id": "2309.01925", "title": "DR-Pose: A Two-stage Deformation-and-Registration Pipeline for\n  Category-level 6D Object Pose Estimation", "abstract": "  Category-level object pose estimation involves estimating the 6D pose and the\n3D metric size of objects from predetermined categories. While recent\napproaches take categorical shape prior information as reference to improve\npose estimation accuracy, the single-stage network design and training manner\nlead to sub-optimal performance since there are two distinct tasks in the\npipeline. In this paper, the advantage of two-stage pipeline over single-stage\ndesign is discussed. To this end, we propose a two-stage deformation-and\nregistration pipeline called DR-Pose, which consists of completion-aided\ndeformation stage and scaled registration stage. The first stage uses a point\ncloud completion method to generate unseen parts of target object, guiding\nsubsequent deformation on the shape prior. In the second stage, a novel\nregistration network is designed to extract pose-sensitive features and predict\nthe representation of object partial point cloud in canonical space based on\nthe deformation results from the first stage. DR-Pose produces superior results\nto the state-of-the-art shape prior-based methods on both CAMERA25 and REAL275\nbenchmarks. Codes are available at https://github.com/Zray26/DR-Pose.git.\n", "date": "2023-09-05", "categories": []}
{"id": "2303.13241", "title": "6D Object Pose Estimation from Approximate 3D Models for Orbital\n  Robotics", "abstract": "  We present a novel technique to estimate the 6D pose of objects from single\nimages where the 3D geometry of the object is only given approximately and not\nas a precise 3D model. To achieve this, we employ a dense 2D-to-3D\ncorrespondence predictor that regresses 3D model coordinates for every pixel.\nIn addition to the 3D coordinates, our model also estimates the pixel-wise\ncoordinate error to discard correspondences that are likely wrong. This allows\nus to generate multiple 6D pose hypotheses of the object, which we then refine\niteratively using a highly efficient region-based approach. We also introduce a\nnovel pixel-wise posterior formulation by which we can estimate the probability\nfor each hypothesis and select the most likely one. As we show in experiments,\nour approach is capable of dealing with extreme visual conditions including\noverexposure, high contrast, or low signal-to-noise ratio. This makes it a\npowerful technique for the particularly challenging task of estimating the pose\nof tumbling satellites for in-orbit robotic applications. Our method achieves\nstate-of-the-art performance on the SPEED+ dataset and has won the SPEC2021\npost-mortem competition.\n", "date": "2023-03-23", "categories": []}
{"id": "2308.16528", "title": "SA6D: Self-Adaptive Few-Shot 6D Pose Estimator for Novel and Occluded\n  Objects", "abstract": "  To enable meaningful robotic manipulation of objects in the real-world, 6D\npose estimation is one of the critical aspects. Most existing approaches have\ndifficulties to extend predictions to scenarios where novel object instances\nare continuously introduced, especially with heavy occlusions. In this work, we\npropose a few-shot pose estimation (FSPE) approach called SA6D, which uses a\nself-adaptive segmentation module to identify the novel target object and\nconstruct a point cloud model of the target object using only a small number of\ncluttered reference images. Unlike existing methods, SA6D does not require\nobject-centric reference images or any additional object information, making it\na more generalizable and scalable solution across categories. We evaluate SA6D\non real-world tabletop object datasets and demonstrate that SA6D outperforms\nexisting FSPE methods, particularly in cluttered scenes with occlusions, while\nrequiring fewer reference images.\n", "date": "2023-08-31", "categories": []}
{"id": "2210.10959", "title": "Geo6D: Geometric Constraints Learning for 6D Pose Estimation", "abstract": "  Numerous 6D pose estimation methods have been proposed that employ end-to-end\nregression to directly estimate the target pose parameters. Since the visible\nfeatures of objects are implicitly influenced by their poses, the network\nallows inferring the pose by analyzing the differences in features in the\nvisible region. However, due to the unpredictable and unrestricted range of\npose variations, the implicitly learned visible feature-pose constraints are\ninsufficiently covered by the training samples, making the network vulnerable\nto unseen object poses. To tackle these challenges, we proposed a novel\ngeometric constraints learning approach called Geo6D for direct regression 6D\npose estimation methods. It introduces a pose transformation formula expressed\nin relative offset representation, which is leveraged as geometric constraints\nto reconstruct the input and output targets of the network. These reconstructed\ndata enable the network to estimate the pose based on explicit geometric\nconstraints and relative offset representation mitigates the issue of the pose\ndistribution gap. Extensive experimental results show that when equipped with\nGeo6D, the direct 6D methods achieve state-of-the-art performance on multiple\ndatasets and demonstrate significant effectiveness, even with only 10% amount\nof data.\n", "date": "2022-10-20", "categories": []}
{"id": "2308.09916", "title": "VI-Net: Boosting Category-level 6D Object Pose Estimation via Learning\n  Decoupled Rotations on the Spherical Representations", "abstract": "  Rotation estimation of high precision from an RGB-D object observation is a\nhuge challenge in 6D object pose estimation, due to the difficulty of learning\nin the non-linear space of SO(3). In this paper, we propose a novel rotation\nestimation network, termed as VI-Net, to make the task easier by decoupling the\nrotation as the combination of a viewpoint rotation and an in-plane rotation.\nMore specifically, VI-Net bases the feature learning on the sphere with two\nindividual branches for the estimates of two factorized rotations, where a\nV-Branch is employed to learn the viewpoint rotation via binary classification\non the spherical signals, while another I-Branch is used to estimate the\nin-plane rotation by transforming the signals to view from the zenith\ndirection. To process the spherical signals, a Spherical Feature Pyramid\nNetwork is constructed based on a novel design of SPAtial Spherical Convolution\n(SPA-SConv), which settles the boundary problem of spherical signals via\nfeature padding and realizesviewpoint-equivariant feature extraction by\nsymmetric convolutional operations. We apply the proposed VI-Net to the\nchallenging task of category-level 6D object pose estimation for predicting the\nposes of unknown objects without available CAD models; experiments on the\nbenchmarking datasets confirm the efficacy of our method, which outperforms the\nexisting ones with a large margin in the regime of high precision.\n", "date": "2023-08-19", "categories": []}
{"id": "2308.10016", "title": "Pseudo Flow Consistency for Self-Supervised 6D Object Pose Estimation", "abstract": "  Most self-supervised 6D object pose estimation methods can only work with\nadditional depth information or rely on the accurate annotation of 2D\nsegmentation masks, limiting their application range. In this paper, we propose\na 6D object pose estimation method that can be trained with pure RGB images\nwithout any auxiliary information. We first obtain a rough pose initialization\nfrom networks trained on synthetic images rendered from the target's 3D mesh.\nThen, we introduce a refinement strategy leveraging the geometry constraint in\nsynthetic-to-real image pairs from multiple different views. We formulate this\ngeometry constraint as pixel-level flow consistency between the training images\nwith dynamically generated pseudo labels. We evaluate our method on three\nchallenging datasets and demonstrate that it outperforms state-of-the-art\nself-supervised methods significantly, with neither 2D annotations nor\nadditional depth images.\n", "date": "2023-08-19", "categories": []}
{"id": "2308.10627", "title": "Polarimetric Information for Multi-Modal 6D Pose Estimation of\n  Photometrically Challenging Objects with Limited Data", "abstract": "  6D pose estimation pipelines that rely on RGB-only or RGB-D data show\nlimitations for photometrically challenging objects with e.g. textureless\nsurfaces, reflections or transparency. A supervised learning-based method\nutilising complementary polarisation information as input modality is proposed\nto overcome such limitations. This supervised approach is then extended to a\nself-supervised paradigm by leveraging physical characteristics of polarised\nlight, thus eliminating the need for annotated real data. The methods achieve\nsignificant advancements in pose estimation by leveraging geometric information\nfrom polarised light and incorporating shape priors and invertible physical\nconstraints.\n", "date": "2023-08-21", "categories": []}
{"id": "2303.16874", "title": "CheckerPose: Progressive Dense Keypoint Localization for Object Pose\n  Estimation with Graph Neural Network", "abstract": "  Estimating the 6-DoF pose of a rigid object from a single RGB image is a\ncrucial yet challenging task. Recent studies have shown the great potential of\ndense correspondence-based solutions, yet improvements are still needed to\nreach practical deployment. In this paper, we propose a novel pose estimation\nalgorithm named CheckerPose, which improves on three main aspects. Firstly,\nCheckerPose densely samples 3D keypoints from the surface of the 3D object and\nfinds their 2D correspondences progressively in the 2D image. Compared to\nprevious solutions that conduct dense sampling in the image space, our strategy\nenables the correspondence searching in a 2D grid (i.e., pixel coordinate).\nSecondly, for our 3D-to-2D correspondence, we design a compact binary code\nrepresentation for 2D image locations. This representation not only allows for\nprogressive correspondence refinement but also converts the correspondence\nregression to a more efficient classification problem. Thirdly, we adopt a\ngraph neural network to explicitly model the interactions among the sampled 3D\nkeypoints, further boosting the reliability and accuracy of the\ncorrespondences. Together, these novel components make CheckerPose a strong\npose estimation algorithm. When evaluated on the popular Linemod, Linemod-O,\nand YCB-V object pose estimation benchmarks, CheckerPose clearly boosts the\naccuracy of correspondence-based methods and achieves state-of-the-art\nperformances. Code is available at https://github.com/RuyiLian/CheckerPose.\n", "date": "2023-03-29", "categories": []}
{"id": "2308.05438", "title": "Deep Fusion Transformer Network with Weighted Vector-Wise Keypoints\n  Voting for Robust 6D Object Pose Estimation", "abstract": "  One critical challenge in 6D object pose estimation from a single RGBD image\nis efficient integration of two different modalities, i.e., color and depth. In\nthis work, we tackle this problem by a novel Deep Fusion Transformer~(DFTr)\nblock that can aggregate cross-modality features for improving pose estimation.\nUnlike existing fusion methods, the proposed DFTr can better model\ncross-modality semantic correlation by leveraging their semantic similarity,\nsuch that globally enhanced features from different modalities can be better\nintegrated for improved information extraction. Moreover, to further improve\nrobustness and efficiency, we introduce a novel weighted vector-wise voting\nalgorithm that employs a non-iterative global optimization strategy for precise\n3D keypoint localization while achieving near real-time inference. Extensive\nexperiments show the effectiveness and strong generalization capability of our\nproposed 3D keypoint voting algorithm. Results on four widely used benchmarks\nalso demonstrate that our method outperforms the state-of-the-art methods by\nlarge margins.\n", "date": "2023-08-10", "categories": []}
{"id": "2308.02239", "title": "DTF-Net: Category-Level Pose Estimation and Shape Reconstruction via\n  Deformable Template Field", "abstract": "  Estimating 6D poses and reconstructing 3D shapes of objects in open-world\nscenes from RGB-depth image pairs is challenging. Many existing methods rely on\nlearning geometric features that correspond to specific templates while\ndisregarding shape variations and pose differences among objects in the same\ncategory. As a result, these methods underperform when handling unseen object\ninstances in complex environments. In contrast, other approaches aim to achieve\ncategory-level estimation and reconstruction by leveraging normalized geometric\nstructure priors, but the static prior-based reconstruction struggles with\nsubstantial intra-class variations. To solve these problems, we propose the\nDTF-Net, a novel framework for pose estimation and shape reconstruction based\non implicit neural fields of object categories. In DTF-Net, we design a\ndeformable template field to represent the general category-wise shape latent\nfeatures and intra-category geometric deformation features. The field\nestablishes continuous shape correspondences, deforming the category template\ninto arbitrary observed instances to accomplish shape reconstruction. We\nintroduce a pose regression module that shares the deformation features and\ntemplate codes from the fields to estimate the accurate 6D pose of each object\nin the scene. We integrate a multi-modal representation extraction module to\nextract object features and semantic masks, enabling end-to-end inference.\nMoreover, during training, we implement a shape-invariant training strategy and\na viewpoint sampling method to further enhance the model's capability to\nextract object pose features. Extensive experiments on the REAL275 and CAMERA25\ndatasets demonstrate the superiority of DTF-Net in both synthetic and real\nscenes. Furthermore, we show that DTF-Net effectively supports grasping tasks\nwith a real robot arm.\n", "date": "2023-08-04", "categories": []}
{"id": "2308.01477", "title": "HANDAL: A Dataset of Real-World Manipulable Object Categories with Pose\n  Annotations, Affordances, and Reconstructions", "abstract": "  We present the HANDAL dataset for category-level object pose estimation and\naffordance prediction. Unlike previous datasets, ours is focused on\nrobotics-ready manipulable objects that are of the proper size and shape for\nfunctional grasping by robot manipulators, such as pliers, utensils, and\nscrewdrivers. Our annotation process is streamlined, requiring only a single\noff-the-shelf camera and semi-automated processing, allowing us to produce\nhigh-quality 3D annotations without crowd-sourcing. The dataset consists of\n308k annotated image frames from 2.2k videos of 212 real-world objects in 17\ncategories. We focus on hardware and kitchen tool objects to facilitate\nresearch in practical scenarios in which a robot manipulator needs to interact\nwith the environment beyond simple pushing or indiscriminate grasping. We\noutline the usefulness of our dataset for 6-DoF category-level pose+scale\nestimation and related tasks. We also provide 3D reconstructed meshes of all\nobjects, and we outline some of the bottlenecks to be addressed for\ndemocratizing the collection of datasets like this one.\n", "date": "2023-08-02", "categories": []}
{"id": "2307.15671", "title": "TrackAgent: 6D Object Tracking via Reinforcement Learning", "abstract": "  Tracking an object's 6D pose, while either the object itself or the observing\ncamera is moving, is important for many robotics and augmented reality\napplications. While exploiting temporal priors eases this problem,\nobject-specific knowledge is required to recover when tracking is lost. Under\nthe tight time constraints of the tracking task, RGB(D)-based methods are often\nconceptionally complex or rely on heuristic motion models. In comparison, we\npropose to simplify object tracking to a reinforced point cloud (depth only)\nalignment task. This allows us to train a streamlined approach from scratch\nwith limited amounts of sparse 3D point clouds, compared to the large datasets\nof diverse RGBD sequences required in previous works. We incorporate temporal\nframe-to-frame registration with object-based recovery by frame-to-model\nrefinement using a reinforcement learning (RL) agent that jointly solves for\nboth objectives. We also show that the RL agent's uncertainty and a\nrendering-based mask propagation are effective reinitialization triggers.\n", "date": "2023-07-28", "categories": []}
{"id": "2307.12400", "title": "TransNet: Transparent Object Manipulation Through Category-Level Pose\n  Estimation", "abstract": "  Transparent objects present multiple distinct challenges to visual perception\nsystems. First, their lack of distinguishing visual features makes transparent\nobjects harder to detect and localize than opaque objects. Even humans find\ncertain transparent surfaces with little specular reflection or refraction,\nlike glass doors, difficult to perceive. A second challenge is that depth\nsensors typically used for opaque object perception cannot obtain accurate\ndepth measurements on transparent surfaces due to their unique reflective\nproperties. Stemming from these challenges, we observe that transparent object\ninstances within the same category, such as cups, look more similar to each\nother than to ordinary opaque objects of that same category. Given this\nobservation, the present paper explores the possibility of category-level\ntransparent object pose estimation rather than instance-level pose estimation.\nWe propose \\textit{\\textbf{TransNet}}, a two-stage pipeline that estimates\ncategory-level transparent object pose using localized depth completion and\nsurface normal estimation. TransNet is evaluated in terms of pose estimation\naccuracy on a large-scale transparent object dataset and compared to a\nstate-of-the-art category-level pose estimation approach. Results from this\ncomparison demonstrate that TransNet achieves improved pose estimation accuracy\non transparent objects. Moreover, we use TransNet to build an autonomous\ntransparent object manipulation system for robotic pick-and-place and pouring\ntasks.\n", "date": "2023-07-23", "categories": []}
{"id": "2302.11827", "title": "Open Challenges for Monocular Single-shot 6D Object Pose Estimation", "abstract": "  Object pose estimation is a non-trivial task that enables robotic\nmanipulation, bin picking, augmented reality, and scene understanding, to name\na few use cases. Monocular object pose estimation gained considerable momentum\nwith the rise of high-performing deep learning-based solutions and is\nparticularly interesting for the community since sensors are inexpensive and\ninference is fast. Prior works establish the comprehensive state of the art for\ndiverse pose estimation problems. Their broad scopes make it difficult to\nidentify promising future directions. We narrow down the scope to the problem\nof single-shot monocular 6D object pose estimation, which is commonly used in\nrobotics, and thus are able to identify such trends. By reviewing recent\npublications in robotics and computer vision, the state of the art is\nestablished at the union of both fields. Following that, we identify promising\nresearch directions in order to help researchers to formulate relevant research\nideas and effectively advance the state of the art. Findings include that\nmethods are sophisticated enough to overcome the domain shift and that\nocclusion handling is a fundamental challenge. We also highlight problems such\nas novel object pose estimation and challenging materials handling as central\nchallenges to advance robotics.\n", "date": "2023-02-23", "categories": []}
{"id": "2307.11550", "title": "YOLOPose V2: Understanding and Improving Transformer-based 6D Pose\n  Estimation", "abstract": "  6D object pose estimation is a crucial prerequisite for autonomous robot\nmanipulation applications. The state-of-the-art models for pose estimation are\nconvolutional neural network (CNN)-based. Lately, Transformers, an architecture\noriginally proposed for natural language processing, is achieving\nstate-of-the-art results in many computer vision tasks as well. Equipped with\nthe multi-head self-attention mechanism, Transformers enable simple\nsingle-stage end-to-end architectures for learning object detection and 6D\nobject pose estimation jointly. In this work, we propose YOLOPose (short form\nfor You Only Look Once Pose estimation), a Transformer-based multi-object 6D\npose estimation method based on keypoint regression and an improved variant of\nthe YOLOPose model. In contrast to the standard heatmaps for predicting\nkeypoints in an image, we directly regress the keypoints. Additionally, we\nemploy a learnable orientation estimation module to predict the orientation\nfrom the keypoints. Along with a separate translation estimation module, our\nmodel is end-to-end differentiable. Our method is suitable for real-time\napplications and achieves results comparable to state-of-the-art methods. We\nanalyze the role of object queries in our architecture and reveal that the\nobject queries specialize in detecting objects in specific image regions.\nFurthermore, we quantify the accuracy trade-off of using datasets of smaller\nsizes to train our model.\n", "date": "2023-07-21", "categories": []}
{"id": "2303.13479", "title": "IST-Net: Prior-free Category-level Pose Estimation with Implicit Space\n  Transformation", "abstract": "  Category-level 6D pose estimation aims to predict the poses and sizes of\nunseen objects from a specific category. Thanks to prior deformation, which\nexplicitly adapts a category-specific 3D prior (i.e., a 3D template) to a given\nobject instance, prior-based methods attained great success and have become a\nmajor research stream. However, obtaining category-specific priors requires\ncollecting a large amount of 3D models, which is labor-consuming and often not\naccessible in practice. This motivates us to investigate whether priors are\nnecessary to make prior-based methods effective. Our empirical study shows that\nthe 3D prior itself is not the credit to the high performance. The keypoint\nactually is the explicit deformation process, which aligns camera and world\ncoordinates supervised by world-space 3D models (also called canonical space).\nInspired by these observations, we introduce a simple prior-free implicit space\ntransformation network, namely IST-Net, to transform camera-space features to\nworld-space counterparts and build correspondence between them in an implicit\nmanner without relying on 3D priors. Besides, we design camera- and world-space\nenhancers to enrich the features with pose-sensitive information and\ngeometrical constraints, respectively. Albeit simple, IST-Net achieves\nstate-of-the-art performance based-on prior-free design, with top inference\nspeed on the REAL275 benchmark. Our code and models are available at\nhttps://github.com/CVMI-Lab/IST-Net.\n", "date": "2023-03-23", "categories": []}
{"id": "2307.05561", "title": "TransPose: A Transformer-based 6D Object Pose Estimation Network with\n  Depth Refinement", "abstract": "  As demand for robotics manipulation application increases, accurate\nvision-based 6D pose estimation becomes essential for autonomous operations.\nConvolutional Neural Networks (CNNs) based approaches for pose estimation have\nbeen previously introduced. However, the quest for better performance still\npersists especially for accurate robotics manipulation. This quest extends to\nthe Agri-robotics domain. In this paper, we propose TransPose, an improved\nTransformer-based 6D pose estimation with a depth refinement module. The\narchitecture takes in only an RGB image as input with no additional\nsupplementing modalities such as depth or thermal images. The architecture\nencompasses an innovative lighter depth estimation network that estimates depth\nfrom an RGB image using feature pyramid with an up-sampling method. A\ntransformer-based detection network with additional prediction heads is\nproposed to directly regress the object's centre and predict the 6D pose of the\ntarget. A novel depth refinement module is then used alongside the predicted\ncenters, 6D poses and depth patches to refine the accuracy of the estimated 6D\npose. We extensively compared our results with other state-of-the-art methods\nand analysed our results for fruit-picking applications. The results we\nachieved show that our proposed technique outperforms the other methods\navailable in the literature.\n", "date": "2023-07-09", "categories": []}
{"id": "2307.03704", "title": "Equivariant Single View Pose Prediction Via Induced and Restricted\n  Representations", "abstract": "  Learning about the three-dimensional world from two-dimensional images is a\nfundamental problem in computer vision. An ideal neural network architecture\nfor such tasks would leverage the fact that objects can be rotated and\ntranslated in three dimensions to make predictions about novel images. However,\nimposing SO(3)-equivariance on two-dimensional inputs is difficult because the\ngroup of three-dimensional rotations does not have a natural action on the\ntwo-dimensional plane. Specifically, it is possible that an element of SO(3)\nwill rotate an image out of plane. We show that an algorithm that learns a\nthree-dimensional representation of the world from two dimensional images must\nsatisfy certain geometric consistency properties which we formulate as\nSO(2)-equivariance constraints. We use the induced and restricted\nrepresentations of SO(2) on SO(3) to construct and classify architectures which\nsatisfy these geometric consistency constraints. We prove that any architecture\nwhich respects said consistency constraints can be realized as an instance of\nour construction. We show that three previously proposed neural architectures\nfor 3D pose prediction are special cases of our construction. We propose a new\nalgorithm that is a learnable generalization of previously considered methods.\nWe test our architecture on three pose predictions task and achieve SOTA\nresults on both the PASCAL3D+ and SYMSOL pose estimation tasks.\n", "date": "2023-07-07", "categories": []}
{"id": "2307.00306", "title": "SyMFM6D: Symmetry-aware Multi-directional Fusion for Multi-View 6D\n  Object Pose Estimation", "abstract": "  Detecting objects and estimating their 6D poses is essential for automated\nsystems to interact safely with the environment. Most 6D pose estimators,\nhowever, rely on a single camera frame and suffer from occlusions and\nambiguities due to object symmetries. We overcome this issue by presenting a\nnovel symmetry-aware multi-view 6D pose estimator called SyMFM6D. Our approach\nefficiently fuses the RGB-D frames from multiple perspectives in a deep\nmulti-directional fusion network and predicts predefined keypoints for all\nobjects in the scene simultaneously. Based on the keypoints and an instance\nsemantic segmentation, we efficiently compute the 6D poses by least-squares\nfitting. To address the ambiguity issues for symmetric objects, we propose a\nnovel training procedure for symmetry-aware keypoint detection including a new\nobjective function. Our SyMFM6D network significantly outperforms the\nstate-of-the-art in both single-view and multi-view 6D pose estimation. We\nfurthermore show the effectiveness of our symmetry-aware training procedure and\ndemonstrate that our approach is robust towards inaccurate camera calibration\nand dynamic camera setups.\n", "date": "2023-07-01", "categories": []}
{"id": "2306.15858", "title": "Hierarchical Graph Neural Networks for Proprioceptive 6D Pose Estimation\n  of In-hand Objects", "abstract": "  Robotic manipulation, in particular in-hand object manipulation, often\nrequires an accurate estimate of the object's 6D pose. To improve the accuracy\nof the estimated pose, state-of-the-art approaches in 6D object pose estimation\nuse observational data from one or more modalities, e.g., RGB images, depth,\nand tactile readings. However, existing approaches make limited use of the\nunderlying geometric structure of the object captured by these modalities,\nthereby, increasing their reliance on visual features. This results in poor\nperformance when presented with objects that lack such visual features or when\nvisual features are simply occluded. Furthermore, current approaches do not\ntake advantage of the proprioceptive information embedded in the position of\nthe fingers. To address these limitations, in this paper: (1) we introduce a\nhierarchical graph neural network architecture for combining multimodal (vision\nand touch) data that allows for a geometrically informed 6D object pose\nestimation, (2) we introduce a hierarchical message passing operation that\nflows the information within and across modalities to learn a graph-based\nobject representation, and (3) we introduce a method that accounts for the\nproprioceptive information for in-hand object representation. We evaluate our\nmodel on a diverse subset of objects from the YCB Object and Model Set, and\nshow that our method substantially outperforms existing state-of-the-art work\nin accuracy and robustness to occlusion. We also deploy our proposed framework\non a real robot and qualitatively demonstrate successful transfer to real\nsettings.\n", "date": "2023-06-28", "categories": []}
{"id": "2306.14056", "title": "Stable Yaw Estimation of Boats from the Viewpoint of UAVs and USVs", "abstract": "  Yaw estimation of boats from the viewpoint of unmanned aerial vehicles (UAVs)\nand unmanned surface vehicles (USVs) or boats is a crucial task in various\napplications such as 3D scene rendering, trajectory prediction, and navigation.\nHowever, the lack of literature on yaw estimation of objects from the viewpoint\nof UAVs has motivated us to address this domain. In this paper, we propose a\nmethod based on HyperPosePDF for predicting the orientation of boats in the 6D\nspace. For that, we use existing datasets, such as PASCAL3D+ and our own\ndatasets, SeaDronesSee-3D and BOArienT, which we annotated manually. We extend\nHyperPosePDF to work in video-based scenarios, such that it yields robust\norientation predictions across time. Naively applying HyperPosePDF on video\ndata yields single-point predictions, resulting in far-off predictions and\noften incorrect symmetric orientations due to unseen or visually different\ndata. To alleviate this issue, we propose aggregating the probability\ndistributions of pose predictions, resulting in significantly improved\nperformance, as shown in our experimental evaluation. Our proposed method could\nsignificantly benefit downstream tasks in marine robotics.\n", "date": "2023-06-24", "categories": []}
{"id": "2306.13266", "title": "Shape-Constraint Recurrent Flow for 6D Object Pose Estimation", "abstract": "  Most recent 6D object pose methods use 2D optical flow to refine their\nresults. However, the general optical flow methods typically do not consider\nthe target's 3D shape information during matching, making them less effective\nin 6D object pose estimation. In this work, we propose a shape-constraint\nrecurrent matching framework for 6D object pose estimation. We first compute a\npose-induced flow based on the displacement of 2D reprojection between the\ninitial pose and the currently estimated pose, which embeds the target's 3D\nshape implicitly. Then we use this pose-induced flow to construct the\ncorrelation map for the following matching iterations, which reduces the\nmatching space significantly and is much easier to learn. Furthermore, we use\nnetworks to learn the object pose based on the current estimated flow, which\nfacilitates the computation of the pose-induced flow for the next iteration and\nyields an end-to-end system for object pose. Finally, we optimize the optical\nflow and object pose simultaneously in a recurrent manner. We evaluate our\nmethod on three challenging 6D object pose datasets and show that it\noutperforms the state of the art significantly in both accuracy and efficiency.\n", "date": "2023-06-23", "categories": []}
{"id": "2306.07598", "title": "Learning to Estimate 6DoF Pose from Limited Data: A Few-Shot,\n  Generalizable Approach using RGB Images", "abstract": "  The accurate estimation of six degrees-of-freedom (6DoF) object poses is\nessential for many applications in robotics and augmented reality. However,\nexisting methods for 6DoF pose estimation often depend on CAD templates or\ndense support views, restricting their usefulness in realworld situations. In\nthis study, we present a new cascade framework named Cas6D for few-shot 6DoF\npose estimation that is generalizable and uses only RGB images. To address the\nfalse positives of target object detection in the extreme few-shot setting, our\nframework utilizes a selfsupervised pre-trained ViT to learn robust feature\nrepresentations. Then, we initialize the nearest top-K pose candidates based on\nsimilarity score and refine the initial poses using feature pyramids to\nformulate and update the cascade warped feature volume, which encodes context\nat increasingly finer scales. By discretizing the pose search range using\nmultiple pose bins and progressively narrowing the pose search range in each\nstage using predictions from the previous stage, Cas6D can overcome the large\ngap between pose candidates and ground truth poses, which is a common failure\nmode in sparse-view scenarios. Experimental results on the LINEMOD and GenMOP\ndatasets demonstrate that Cas6D outperforms state-of-the-art methods by 9.2%\nand 3.8% accuracy (Proj-5) under the 32-shot setting compared to OnePose++ and\nGen6D.\n", "date": "2023-06-13", "categories": []}
{"id": "2306.00118", "title": "Neural Textured Deformable Meshes for Robust Analysis-by-Synthesis", "abstract": "  Human vision demonstrates higher robustness than current AI algorithms under\nout-of-distribution scenarios. It has been conjectured such robustness benefits\nfrom performing analysis-by-synthesis. Our paper formulates triple vision tasks\nin a consistent manner using approximate analysis-by-synthesis by\nrender-and-compare algorithms on neural features. In this work, we introduce\nNeural Textured Deformable Meshes, which involve the object model with\ndeformable geometry that allows optimization on both camera parameters and\nobject geometries. The deformable mesh is parameterized as a neural field, and\ncovered by whole-surface neural texture maps, which are trained to have spatial\ndiscriminability. During inference, we extract the feature map of the test\nimage and subsequently optimize the 3D pose and shape parameters of our model\nusing differentiable rendering to best reconstruct the target feature map. We\nshow that our analysis-by-synthesis is much more robust than conventional\nneural networks when evaluated on real-world images and even in challenging\nout-of-distribution scenarios, such as occlusion and domain shift. Our\nalgorithms are competitive with standard algorithms when tested on conventional\nperformance measures.\n", "date": "2023-05-31", "categories": []}
{"id": "2306.00129", "title": "Self-supervised Vision Transformers for 3D Pose Estimation of Novel\n  Objects", "abstract": "  Object pose estimation is important for object manipulation and scene\nunderstanding. In order to improve the general applicability of pose\nestimators, recent research focuses on providing estimates for novel objects,\nthat is objects unseen during training. Such works use deep template matching\nstrategies to retrieve the closest template connected to a query image. This\ntemplate retrieval implicitly provides object class and pose. Despite the\nrecent success and improvements of Vision Transformers over CNNs for many\nvision tasks, the state of the art uses CNN-based approaches for novel object\npose estimation. This work evaluates and demonstrates the differences between\nself-supervised CNNs and Vision Transformers for deep template matching. In\ndetail, both types of approaches are trained using contrastive learning to\nmatch training images against rendered templates of isolated objects. At test\ntime, such templates are matched against query images of known and novel\nobjects under challenging settings, such as clutter, occlusion and object\nsymmetries, using masked cosine similarity. The presented results not only\ndemonstrate that Vision Transformers improve in matching accuracy over CNNs,\nbut also that for some cases pre-trained Vision Transformers do not need\nfine-tuning to do so. Furthermore, we highlight the differences in optimization\nand network architecture when comparing these two types of network for deep\ntemplate matching.\n", "date": "2023-05-31", "categories": []}
{"id": "2306.00892", "title": "A Probabilistic Relaxation of the Two-Stage Object Pose Estimation\n  Paradigm", "abstract": "  Existing object pose estimation methods commonly require a one-to-one point\nmatching step that forces them to be separated into two consecutive stages:\nvisual correspondence detection (e.g., by matching feature descriptors as part\nof a perception front-end) followed by geometric alignment (e.g., by optimizing\na robust estimation objective for pointcloud registration or\nperspective-n-point). Instead, we propose a matching-free probabilistic\nformulation with two main benefits: i) it enables unified and concurrent\noptimization of both visual correspondence and geometric alignment, and ii) it\ncan represent different plausible modes of the entire distribution of likely\nposes. This in turn allows for a more graceful treatment of geometric\nperception scenarios where establishing one-to-one matches between points is\nconceptually ill-defined, such as textureless, symmetrical and/or occluded\nobjects and scenes where the correct pose is uncertain or there are multiple\nequally valid solutions.\n", "date": "2023-06-01", "categories": []}
{"id": "2305.17252", "title": "Generalizable Pose Estimation Using Implicit Scene Representations", "abstract": "  6-DoF pose estimation is an essential component of robotic manipulation\npipelines. However, it usually suffers from a lack of generalization to new\ninstances and object types. Most widely used methods learn to infer the object\npose in a discriminative setup where the model filters useful information to\ninfer the exact pose of the object. While such methods offer accurate poses,\nthe model does not store enough information to generalize to new objects. In\nthis work, we address the generalization capability of pose estimation using\nmodels that contain enough information about the object to render it in\ndifferent poses. We follow the line of work that inverts neural renderers to\ninfer the pose. We propose i-$\\sigma$SRN to maximize the information flowing\nfrom the input pose to the rendered scene and invert them to infer the pose\ngiven an input image. Specifically, we extend Scene Representation Networks\n(SRNs) by incorporating a separate network for density estimation and introduce\na new way of obtaining a weighted scene representation. We investigate several\nways of initial pose estimates and losses for the neural renderer. Our final\nevaluation shows a significant improvement in inference performance and speed\ncompared to existing approaches.\n", "date": "2023-05-26", "categories": []}
{"id": "2305.15727", "title": "POPE: 6-DoF Promptable Pose Estimation of Any Object, in Any Scene, with\n  One Reference", "abstract": "  Despite the significant progress in six degrees-of-freedom (6DoF) object pose\nestimation, existing methods have limited applicability in real-world scenarios\ninvolving embodied agents and downstream 3D vision tasks. These limitations\nmainly come from the necessity of 3D models, closed-category detection, and a\nlarge number of densely annotated support views. To mitigate this issue, we\npropose a general paradigm for object pose estimation, called Promptable Object\nPose Estimation (POPE). The proposed approach POPE enables zero-shot 6DoF\nobject pose estimation for any target object in any scene, while only a single\nreference is adopted as the support view. To achieve this, POPE leverages the\npower of the pre-trained large-scale 2D foundation model, employs a framework\nwith hierarchical feature representation and 3D geometry principles. Moreover,\nit estimates the relative camera pose between object prompts and the target\nobject in new views, enabling both two-view and multi-view 6DoF pose estimation\ntasks. Comprehensive experimental results demonstrate that POPE exhibits\nunrivaled robust performance in zero-shot settings, by achieving a significant\nreduction in the averaged Median Pose Error by 52.38% and 50.47% on the LINEMOD\nand OnePose datasets, respectively. We also conduct more challenging testings\nin causally captured images (see Figure 1), which further demonstrates the\nrobustness of POPE. Project page can be found with\nhttps://paulpanwang.github.io/POPE/.\n", "date": "2023-05-25", "categories": []}
{"id": "2305.16124", "title": "Robust Category-Level 3D Pose Estimation from Synthetic Data", "abstract": "  Obtaining accurate 3D object poses is vital for numerous computer vision\napplications, such as 3D reconstruction and scene understanding. However,\nannotating real-world objects is time-consuming and challenging. While\nsynthetically generated training data is a viable alternative, the domain shift\nbetween real and synthetic data is a significant challenge. In this work, we\naim to narrow the performance gap between models trained on synthetic data and\nfew real images and fully supervised models trained on large-scale data. We\nachieve this by approaching the problem from two perspectives: 1) We introduce\nSyntheticP3D, a new synthetic dataset for object pose estimation generated from\nCAD models and enhanced with a novel algorithm. 2) We propose a novel approach\n(CC3D) for training neural mesh models that perform pose estimation via inverse\nrendering. In particular, we exploit the spatial relationships between features\non the mesh surface and a contrastive learning scheme to guide the domain\nadaptation process. Combined, these two approaches enable our models to perform\ncompetitively with state-of-the-art models using only 10% of the respective\nreal training images, while outperforming the SOTA model by 10.4% with a\nthreshold of pi/18 using only 50% of the real training data. Our trained model\nfurther demonstrates robust generalization to out-of-distribution scenarios\ndespite being trained with minimal real data.\n", "date": "2023-05-25", "categories": []}
{"id": "2305.12626", "title": "You Only Look at One: Category-Level Object Representations for Pose\n  Estimation From a Single Example", "abstract": "  In order to meaningfully interact with the world, robot manipulators must be\nable to interpret objects they encounter. A critical aspect of this\ninterpretation is pose estimation: inferring quantities that describe the\nposition and orientation of an object in 3D space. Most existing approaches to\npose estimation make limiting assumptions, often working only for specific,\nknown object instances, or at best generalising to an object category using\nlarge pose-labelled datasets. In this work, we present a method for achieving\ncategory-level pose estimation by inspection of just a single object from a\ndesired category. We show that we can subsequently perform accurate pose\nestimation for unseen objects from an inspected category, and considerably\noutperform prior work by exploiting multi-view correspondences. We demonstrate\nthat our method runs in real-time, enabling a robot manipulator equipped with\nan RGBD sensor to perform online 6D pose estimation for novel objects. Finally,\nwe showcase our method in a continual learning setting, with a robot able to\ndetermine whether objects belong to known categories, and if not, use active\nperception to produce a one-shot category representation for subsequent pose\nestimation.\n", "date": "2023-05-22", "categories": []}
{"id": "2302.06019", "title": "A Correct-and-Certify Approach to Self-Supervise Object Pose Estimators\n  via Ensemble Self-Training", "abstract": "  Real-world robotics applications demand object pose estimation methods that\nwork reliably across a variety of scenarios. Modern learning-based approaches\nrequire large labeled datasets and tend to perform poorly outside the training\ndomain. Our first contribution is to develop a robust corrector module that\ncorrects pose estimates using depth information, thus enabling existing methods\nto better generalize to new test domains; the corrector operates on semantic\nkeypoints (but is also applicable to other pose estimators) and is fully\ndifferentiable. Our second contribution is an ensemble self-training approach\nthat simultaneously trains multiple pose estimators in a self-supervised\nmanner. Our ensemble self-training architecture uses the robust corrector to\nrefine the output of each pose estimator; then, it evaluates the quality of the\noutputs using observable correctness certificates; finally, it uses the\nobservably correct outputs for further training, without requiring external\nsupervision. As an additional contribution, we propose small improvements to a\nregression-based keypoint detection architecture, to enhance its robustness to\noutliers; these improvements include a robust pooling scheme and a robust\ncentroid computation. Experiments on the YCBV and TLESS datasets show the\nproposed ensemble self-training outperforms fully supervised baselines while\nnot requiring 3D annotations on real data.\n", "date": "2023-02-12", "categories": []}
{"id": "2206.11215", "title": "Certifiable 3D Object Pose Estimation: Foundations, Learning Models, and\n  Self-Training", "abstract": "  We consider a certifiable object pose estimation problem, where -- given a\npartial point cloud of an object -- the goal is to not only estimate the object\npose, but also to provide a certificate of correctness for the resulting\nestimate. Our first contribution is a general theory of certification for\nend-to-end perception models. In particular, we introduce the notion of\n$\\zeta$-correctness, which bounds the distance between an estimate and the\nground truth. We show that $\\zeta$-correctness can be assessed by implementing\ntwo certificates: (i) a certificate of observable correctness, that asserts if\nthe model output is consistent with the input data and prior information, (ii)\na certificate of non-degeneracy, that asserts whether the input data is\nsufficient to compute a unique estimate. Our second contribution is to apply\nthis theory and design a new learning-based certifiable pose estimator. We\npropose C-3PO, a semantic-keypoint-based pose estimation model, augmented with\nthe two certificates, to solve the certifiable pose estimation problem. C-3PO\nalso includes a keypoint corrector, implemented as a differentiable\noptimization layer, that can correct large detection errors (e.g. due to the\nsim-to-real gap). Our third contribution is a novel self-supervised training\napproach that uses our certificate of observable correctness to provide the\nsupervisory signal to C-3PO during training. In it, the model trains only on\nthe observably correct input-output pairs, in each training iteration. As\ntraining progresses, we see that the observably correct input-output pairs\ngrow, eventually reaching near 100% in many cases. Our experiments show that\n(i) standard semantic-keypoint-based methods outperform more recent\nalternatives, (ii) C-3PO further improves performance and significantly\noutperforms all the baselines, and (iii) C-3PO's certificates are able to\ndiscern correct pose estimates.\n", "date": "2022-06-22", "categories": []}
{"id": "2303.02133", "title": "Depth-based 6DoF Object Pose Estimation using Swin Transformer", "abstract": "  Accurately estimating the 6D pose of objects is crucial for many\napplications, such as robotic grasping, autonomous driving, and augmented\nreality. However, this task becomes more challenging in poor lighting\nconditions or when dealing with textureless objects. To address this issue,\ndepth images are becoming an increasingly popular choice due to their\ninvariance to a scene's appearance and the implicit incorporation of essential\ngeometric characteristics. However, fully leveraging depth information to\nimprove the performance of pose estimation remains a difficult and\nunder-investigated problem. To tackle this challenge, we propose a novel\nframework called SwinDePose, that uses only geometric information from depth\nimages to achieve accurate 6D pose estimation. SwinDePose first calculates the\nangles between each normal vector defined in a depth image and the three\ncoordinate axes in the camera coordinate system. The resulting angles are then\nformed into an image, which is encoded using Swin Transformer. Additionally, we\napply RandLA-Net to learn the representations from point clouds. The resulting\nimage and point clouds embeddings are concatenated and fed into a semantic\nsegmentation module and a 3D keypoints localization module. Finally, we\nestimate 6D poses using a least-square fitting approach based on the target\nobject's predicted semantic mask and 3D keypoints. In experiments on the\nLineMod and Occlusion LineMod datasets, SwinDePose outperforms existing\nstate-of-the-art methods for 6D object pose estimation using depth images. This\ndemonstrates the effectiveness of our approach and highlights its potential for\nimproving performance in real-world scenarios. Our code is at\nhttps://github.com/zhujunli1993/SwinDePose.\n", "date": "2023-03-03", "categories": []}
{"id": "2304.08230", "title": "Uncovering the Background-Induced bias in RGB based 6-DoF Object Pose\n  Estimation", "abstract": "  In recent years, there has been a growing trend of using data-driven methods\nin industrial settings. These kinds of methods often process video images or\nparts, therefore the integrity of such images is crucial. Sometimes datasets,\ne.g. consisting of images, can be sophisticated for various reasons. It becomes\ncritical to understand how the manipulation of video and images can impact the\neffectiveness of a machine learning method. Our case study aims precisely to\nanalyze the Linemod dataset, considered the state of the art in 6D pose\nestimation context. That dataset presents images accompanied by ArUco markers;\nit is evident that such markers will not be available in real-world contexts.\nWe analyze how the presence of the markers affects the pose estimation\naccuracy, and how this bias may be mitigated through data augmentation and\nother methods. Our work aims to show how the presence of these markers goes to\nmodify, in the testing phase, the effectiveness of the deep learning method\nused. In particular, we will demonstrate, through the tool of saliency maps,\nhow the focus of the neural network is captured in part by these ArUco markers.\nFinally, a new dataset, obtained by applying geometric tools to Linemod, will\nbe proposed in order to demonstrate our hypothesis and uncovering the bias. Our\nresults demonstrate the potential for bias in 6DOF pose estimation networks,\nand suggest methods for reducing this bias when training with markers.\n", "date": "2023-04-17", "categories": []}
{"id": "2212.01920", "title": "Review on 6D Object Pose Estimation with the focus on Indoor Scene\n  Understanding", "abstract": "  6D object pose estimation problem has been extensively studied in the field\nof Computer Vision and Robotics. It has wide range of applications such as\nrobot manipulation, augmented reality, and 3D scene understanding. With the\nadvent of Deep Learning, many breakthroughs have been made; however, approaches\ncontinue to struggle when they encounter unseen instances, new categories, or\nreal-world challenges such as cluttered backgrounds and occlusions. In this\nstudy, we will explore the available methods based on input modality, problem\nformulation, and whether it is a category-level or instance-level approach. As\na part of our discussion, we will focus on how 6D object pose estimation can be\nused for understanding 3D scenes.\n", "date": "2022-12-04", "categories": []}
{"id": "2209.07976", "title": "Imitrob: Imitation Learning Dataset for Training and Evaluating 6D\n  Object Pose Estimators", "abstract": "  This paper introduces a dataset for training and evaluating methods for 6D\npose estimation of hand-held tools in task demonstrations captured by a\nstandard RGB camera. Despite the significant progress of 6D pose estimation\nmethods, their performance is usually limited for heavily occluded objects,\nwhich is a common case in imitation learning, where the object is typically\npartially occluded by the manipulating hand. Currently, there is a lack of\ndatasets that would enable the development of robust 6D pose estimation methods\nfor these conditions. To overcome this problem, we collect a new dataset\n(Imitrob) aimed at 6D pose estimation in imitation learning and other\napplications where a human holds a tool and performs a task. The dataset\ncontains image sequences of nine different tools and twelve manipulation tasks\nwith two camera viewpoints, four human subjects, and left/right hand. Each\nimage is accompanied by an accurate ground truth measurement of the 6D object\npose obtained by the HTC Vive motion tracking device. The use of the dataset is\ndemonstrated by training and evaluating a recent 6D object pose estimation\nmethod (DOPE) in various setups.\n", "date": "2022-09-16", "categories": []}
{"id": "2304.01382", "title": "PoseMatcher: One-shot 6D Object Pose Estimation by Deep Feature Matching", "abstract": "  Estimating the pose of an unseen object is the goal of the challenging\none-shot pose estimation task. Previous methods have heavily relied on feature\nmatching with great success. However, these methods are often inefficient and\nlimited by their reliance on pre-trained models that have not be designed\nspecifically for pose estimation. In this paper we propose PoseMatcher, an\naccurate model free one-shot object pose estimator that overcomes these\nlimitations. We create a new training pipeline for object to image matching\nbased on a three-view system: a query with a positive and negative templates.\nThis simple yet effective approach emulates test time scenarios by cheaply\nconstructing an approximation of the full object point cloud during training.\nTo enable PoseMatcher to attend to distinct input modalities, an image and a\npointcloud, we introduce IO-Layer, a new attention layer that efficiently\naccommodates self and cross attention between the inputs. Moreover, we propose\na pruning strategy where we iteratively remove redundant regions of the target\nobject to further reduce the complexity and noise of the network while\nmaintaining accuracy. Finally we redesign commonly used pose refinement\nstrategies, zoom and 2D offset refinements, and adapt them to the one-shot\nparadigm. We outperform all prior one-shot pose estimation methods on the\nLinemod and YCB-V datasets as well achieve results rivaling recent\ninstance-level methods. The source code and models are available at\nhttps://github.com/PedroCastro/PoseMatcher.\n", "date": "2023-04-03", "categories": []}
{"id": "2210.07199", "title": "Self-Supervised Geometric Correspondence for Category-Level 6D Object\n  Pose Estimation in the Wild", "abstract": "  While 6D object pose estimation has wide applications across computer vision\nand robotics, it remains far from being solved due to the lack of annotations.\nThe problem becomes even more challenging when moving to category-level 6D\npose, which requires generalization to unseen instances. Current approaches are\nrestricted by leveraging annotations from simulation or collected from humans.\nIn this paper, we overcome this barrier by introducing a self-supervised\nlearning approach trained directly on large-scale real-world object videos for\ncategory-level 6D pose estimation in the wild. Our framework reconstructs the\ncanonical 3D shape of an object category and learns dense correspondences\nbetween input images and the canonical shape via surface embedding. For\ntraining, we propose novel geometrical cycle-consistency losses which construct\ncycles across 2D-3D spaces, across different instances and different time\nsteps. The learned correspondence can be applied for 6D pose estimation and\nother downstream tasks such as keypoint transfer. Surprisingly, our method,\nwithout any human annotations or simulators, can achieve on-par or even better\nperformance than previous supervised or semi-supervised methods on in-the-wild\nimages. Our project page is: https://kywind.github.io/self-pose .\n", "date": "2022-10-13", "categories": []}
{"id": "2303.16730", "title": "TTA-COPE: Test-Time Adaptation for Category-Level Object Pose Estimation", "abstract": "  Test-time adaptation methods have been gaining attention recently as a\npractical solution for addressing source-to-target domain gaps by gradually\nupdating the model without requiring labels on the target data. In this paper,\nwe propose a method of test-time adaptation for category-level object pose\nestimation called TTA-COPE. We design a pose ensemble approach with a\nself-training loss using pose-aware confidence. Unlike previous unsupervised\ndomain adaptation methods for category-level object pose estimation, our\napproach processes the test data in a sequential, online manner, and it does\nnot require access to the source domain at runtime. Extensive experimental\nresults demonstrate that the proposed pose ensemble and the self-training loss\nimprove category-level object pose performance during test time under both\nsemi-supervised and unsupervised settings. Project page:\nhttps://taeyeop.com/ttacope\n", "date": "2023-03-29", "categories": []}
{"id": "2303.16833", "title": "Multi-View Keypoints for Reliable 6D Object Pose Estimation", "abstract": "  6D Object pose estimation is a fundamental component in robotics enabling\nefficient interaction with the environment. It is particularly challenging in\nbin-picking applications, where many objects are low-feature and reflective,\nand self-occlusion between objects of the same type is common. We propose a\nnovel multi-view approach leveraging known camera transformations from an\neye-in-hand setup to combine heatmap and keypoint estimates into a probability\ndensity map over 3D space. The result is a robust approach that is scalable in\nthe number of views. It relies on a confidence score composed of keypoint\nprobabilities and point-cloud alignment error, which allows reliable rejection\nof false positives. We demonstrate an average pose estimation error of\napproximately 0.5mm and 2 degrees across a variety of difficult low-feature and\nreflective objects in the ROBI dataset, while also surpassing the state-of-art\ncorrect detection rate, measured using the 10% object diameter threshold on ADD\nerror.\n", "date": "2023-03-29", "categories": []}
{"id": "2303.15743", "title": "HS-Pose: Hybrid Scope Feature Extraction for Category-level Object Pose\n  Estimation", "abstract": "  In this paper, we focus on the problem of category-level object pose\nestimation, which is challenging due to the large intra-category shape\nvariation. 3D graph convolution (3D-GC) based methods have been widely used to\nextract local geometric features, but they have limitations for complex shaped\nobjects and are sensitive to noise. Moreover, the scale and translation\ninvariant properties of 3D-GC restrict the perception of an object's size and\ntranslation information. In this paper, we propose a simple network structure,\nthe HS-layer, which extends 3D-GC to extract hybrid scope latent features from\npoint cloud data for category-level object pose estimation tasks. The proposed\nHS-layer: 1) is able to perceive local-global geometric structure and global\ninformation, 2) is robust to noise, and 3) can encode size and translation\ninformation. Our experiments show that the simple replacement of the 3D-GC\nlayer with the proposed HS-layer on the baseline method (GPV-Pose) achieves a\nsignificant improvement, with the performance increased by 14.5% on 5d2cm\nmetric and 10.3% on IoU75. Our method outperforms the state-of-the-art methods\nby a large margin (8.3% on 5d2cm, 6.9% on IoU75) on the REAL275 dataset and\nruns in real-time (50 FPS).\n", "date": "2023-03-28", "categories": []}
{"id": "2303.14158", "title": "BundleSDF: Neural 6-DoF Tracking and 3D Reconstruction of Unknown\n  Objects", "abstract": "  We present a near real-time method for 6-DoF tracking of an unknown object\nfrom a monocular RGBD video sequence, while simultaneously performing neural 3D\nreconstruction of the object. Our method works for arbitrary rigid objects,\neven when visual texture is largely absent. The object is assumed to be\nsegmented in the first frame only. No additional information is required, and\nno assumption is made about the interaction agent. Key to our method is a\nNeural Object Field that is learned concurrently with a pose graph optimization\nprocess in order to robustly accumulate information into a consistent 3D\nrepresentation capturing both geometry and appearance. A dynamic pool of posed\nmemory frames is automatically maintained to facilitate communication between\nthese threads. Our approach handles challenging sequences with large pose\nchanges, partial and full occlusion, untextured surfaces, and specular\nhighlights. We show results on HO3D, YCBInEOAT, and BEHAVE datasets,\ndemonstrating that our method significantly outperforms existing approaches.\nProject page: https://bundlesdf.github.io\n", "date": "2023-03-24", "categories": []}
{"id": "2210.00924", "title": "Multi-view object pose estimation from correspondence distributions and\n  epipolar geometry", "abstract": "  In many automation tasks involving manipulation of rigid objects, the poses\nof the objects must be acquired. Vision-based pose estimation using a single\nRGB or RGB-D sensor is especially popular due to its broad applicability.\nHowever, single-view pose estimation is inherently limited by depth ambiguity\nand ambiguities imposed by various phenomena like occlusion, self-occlusion,\nreflections, etc. Aggregation of information from multiple views can\npotentially resolve these ambiguities, but the current state-of-the-art\nmulti-view pose estimation method only uses multiple views to aggregate\nsingle-view pose estimates, and thus rely on obtaining good single-view\nestimates. We present a multi-view pose estimation method which aggregates\nlearned 2D-3D distributions from multiple views for both the initial estimate\nand optional refinement. Our method performs probabilistic sampling of 3D-3D\ncorrespondences under epipolar constraints using learned 2D-3D correspondence\ndistributions which are implicitly trained to respect visual ambiguities such\nas symmetry. Evaluation on the T-LESS dataset shows that our method reduces\npose estimation errors by 80-91% compared to the best single-view method, and\nwe present state-of-the-art results on T-LESS with four views, even compared\nwith methods using five and eight views.\n", "date": "2022-10-03", "categories": []}
{"id": "2303.13477", "title": "TransPoser: Transformer as an Optimizer for Joint Object Shape and Pose\n  Estimation", "abstract": "  We propose a novel method for joint estimation of shape and pose of rigid\nobjects from their sequentially observed RGB-D images. In sharp contrast to\npast approaches that rely on complex non-linear optimization, we propose to\nformulate it as a neural optimization that learns to efficiently estimate the\nshape and pose. We introduce Deep Directional Distance Function (DeepDDF), a\nneural network that directly outputs the depth image of an object given the\ncamera viewpoint and viewing direction, for efficient error computation in 2D\nimage space. We formulate the joint estimation itself as a Transformer which we\nrefer to as TransPoser. We fully leverage the tokenization and multi-head\nattention to sequentially process the growing set of observations and to\nefficiently update the shape and pose with a learned momentum, respectively.\nExperimental results on synthetic and real data show that DeepDDF achieves high\naccuracy as a category-level object shape representation and TransPoser\nachieves state-of-the-art accuracy efficiently for joint shape and pose\nestimation.\n", "date": "2023-03-23", "categories": []}
{"id": "2303.12246", "title": "Object Pose Estimation with Statistical Guarantees: Conformal Keypoint\n  Detection and Geometric Uncertainty Propagation", "abstract": "  The two-stage object pose estimation paradigm first detects semantic\nkeypoints on the image and then estimates the 6D pose by minimizing\nreprojection errors. Despite performing well on standard benchmarks, existing\ntechniques offer no provable guarantees on the quality and uncertainty of the\nestimation. In this paper, we inject two fundamental changes, namely conformal\nkeypoint detection and geometric uncertainty propagation, into the two-stage\nparadigm and propose the first pose estimator that endows an estimation with\nprovable and computable worst-case error bounds. On one hand, conformal\nkeypoint detection applies the statistical machinery of inductive conformal\nprediction to convert heuristic keypoint detections into circular or elliptical\nprediction sets that cover the groundtruth keypoints with a user-specified\nmarginal probability (e.g., 90%). Geometric uncertainty propagation, on the\nother, propagates the geometric constraints on the keypoints to the 6D object\npose, leading to a Pose UnceRtainty SEt (PURSE) that guarantees coverage of the\ngroundtruth pose with the same probability. The PURSE, however, is a nonconvex\nset that does not directly lead to estimated poses and uncertainties.\nTherefore, we develop RANdom SAmple averaGing (RANSAG) to compute an average\npose and apply semidefinite relaxation to upper bound the worst-case errors\nbetween the average pose and the groundtruth. On the LineMOD Occlusion dataset\nwe demonstrate: (i) the PURSE covers the groundtruth with valid probabilities;\n(ii) the worst-case error bounds provide correct uncertainty quantification;\nand (iii) the average pose achieves better or similar accuracy as\nrepresentative methods based on sparse keypoints.\n", "date": "2023-03-22", "categories": []}
{"id": "2303.12396", "title": "Rigidity-Aware Detection for 6D Object Pose Estimation", "abstract": "  Most recent 6D object pose estimation methods first use object detection to\nobtain 2D bounding boxes before actually regressing the pose. However, the\ngeneral object detection methods they use are ill-suited to handle cluttered\nscenes, thus producing poor initialization to the subsequent pose network. To\naddress this, we propose a rigidity-aware detection method exploiting the fact\nthat, in 6D pose estimation, the target objects are rigid. This lets us\nintroduce an approach to sampling positive object regions from the entire\nvisible object area during training, instead of naively drawing samples from\nthe bounding box center where the object might be occluded. As such, every\nvisible object part can contribute to the final bounding box prediction,\nyielding better detection robustness. Key to the success of our approach is a\nvisibility map, which we propose to build using a minimum barrier distance\nbetween every pixel in the bounding box and the box boundary. Our results on\nseven challenging 6D pose estimation datasets evidence that our method\noutperforms general detection frameworks by a large margin. Furthermore,\ncombined with a pose regression network, we obtain state-of-the-art pose\nestimation results on the challenging BOP benchmark.\n", "date": "2023-03-22", "categories": []}
{"id": "2302.14332", "title": "Markerless Camera-to-Robot Pose Estimation via Self-supervised\n  Sim-to-Real Transfer", "abstract": "  Solving the camera-to-robot pose is a fundamental requirement for\nvision-based robot control, and is a process that takes considerable effort and\ncares to make accurate. Traditional approaches require modification of the\nrobot via markers, and subsequent deep learning approaches enabled markerless\nfeature extraction. Mainstream deep learning methods only use synthetic data\nand rely on Domain Randomization to fill the sim-to-real gap, because acquiring\nthe 3D annotation is labor-intensive. In this work, we go beyond the limitation\nof 3D annotations for real-world data. We propose an end-to-end pose estimation\nframework that is capable of online camera-to-robot calibration and a\nself-supervised training method to scale the training to unlabeled real-world\ndata. Our framework combines deep learning and geometric vision for solving the\nrobot pose, and the pipeline is fully differentiable. To train the\nCamera-to-Robot Pose Estimation Network (CtRNet), we leverage foreground\nsegmentation and differentiable rendering for image-level self-supervision. The\npose prediction is visualized through a renderer and the image loss with the\ninput image is back-propagated to train the neural network. Our experimental\nresults on two public real datasets confirm the effectiveness of our approach\nover existing works. We also integrate our framework into a visual servoing\nsystem to demonstrate the promise of real-time precise robot pose estimation\nfor automation tasks.\n", "date": "2023-02-28", "categories": []}
{"id": "2209.09659", "title": "Ki-Pode: Keypoint-based Implicit Pose Distribution Estimation of Rigid\n  Objects", "abstract": "  The estimation of 6D poses of rigid objects is a fundamental problem in\ncomputer vision. Traditionally pose estimation is concerned with the\ndetermination of a single best estimate. However, a single estimate is unable\nto express visual ambiguity, which in many cases is unavoidable due to object\nsymmetries or occlusion of identifying features. Inability to account for\nambiguities in pose can lead to failure in subsequent methods, which is\nunacceptable when the cost of failure is high. Estimates of full pose\ndistributions are, contrary to single estimates, well suited for expressing\nuncertainty on pose. Motivated by this, we propose a novel pose distribution\nestimation method. An implicit formulation of the probability distribution over\nobject pose is derived from an intermediary representation of an object as a\nset of keypoints. This ensures that the pose distribution estimates have a high\nlevel of interpretability. Furthermore, our method is based on conservative\napproximations, which leads to reliable estimates. The method has been\nevaluated on the task of rotation distribution estimation on the YCB-V and\nT-LESS datasets and performs reliably on all objects.\n", "date": "2022-09-20", "categories": []}
{"id": "2303.10346", "title": "SOCS: Semantically-aware Object Coordinate Space for Category-Level 6D\n  Object Pose Estimation under Large Shape Variations", "abstract": "  Most learning-based approaches to category-level 6D pose estimation are\ndesign around normalized object coordinate space (NOCS). While being\nsuccessful, NOCS-based methods become inaccurate and less robust when handling\nobjects of a category containing significant intra-category shape variations.\nThis is because the object coordinates induced by global and rigid alignment of\nobjects are semantically incoherent, making the coordinate regression hard to\nlearn and generalize. We propose Semantically-aware Object Coordinate Space\n(SOCS) built by warping-and-aligning the objects guided by a sparse set of\nkeypoints with semantically meaningful correspondence. SOCS is semantically\ncoherent: Any point on the surface of a object can be mapped to a semantically\nmeaningful location in SOCS, allowing for accurate pose and size estimation\nunder large shape variations. To learn effective coordinate regression to SOCS,\nwe propose a novel multi-scale coordinate-based attention network. Evaluations\ndemonstrate that our method is easy to train, well-generalizing for large\nintra-category shape variations and robust to inter-object occlusions.\n", "date": "2023-03-18", "categories": []}
{"id": "2208.06416", "title": "Uni6Dv2: Noise Elimination for 6D Pose Estimation", "abstract": "  Uni6D is the first 6D pose estimation approach to employ a unified backbone\nnetwork to extract features from both RGB and depth images. We discover that\nthe principal reasons of Uni6D performance limitations are Instance-Outside and\nInstance-Inside noise. Uni6D's simple pipeline design inherently introduces\nInstance-Outside noise from background pixels in the receptive field, while\nignoring Instance-Inside noise in the input depth data. In this paper, we\npropose a two-step denoising approach for dealing with the aforementioned noise\nin Uni6D. To reduce noise from non-instance regions, an instance segmentation\nnetwork is utilized in the first step to crop and mask the instance. A\nlightweight depth denoising module is proposed in the second step to calibrate\nthe depth feature before feeding it into the pose regression network. Extensive\nexperiments show that our Uni6Dv2 reliably and robustly eliminates noise,\noutperforming Uni6D without sacrificing too much inference efficiency. It also\nreduces the need for annotated real data that requires costly labeling.\n", "date": "2022-08-15", "categories": []}
{"id": "2212.12902", "title": "TexPose: Neural Texture Learning for Self-Supervised 6D Object Pose\n  Estimation", "abstract": "  In this paper, we introduce neural texture learning for 6D object pose\nestimation from synthetic data and a few unlabelled real images. Our major\ncontribution is a novel learning scheme which removes the drawbacks of previous\nworks, namely the strong dependency on co-modalities or additional refinement.\nThese have been previously necessary to provide training signals for\nconvergence. We formulate such a scheme as two sub-optimisation problems on\ntexture learning and pose learning. We separately learn to predict realistic\ntexture of objects from real image collections and learn pose estimation from\npixel-perfect synthetic data. Combining these two capabilities allows then to\nsynthesise photorealistic novel views to supervise the pose estimator with\naccurate geometry. To alleviate pose noise and segmentation imperfection\npresent during the texture learning phase, we propose a surfel-based\nadversarial training loss together with texture regularisation from synthetic\ndata. We demonstrate that the proposed approach significantly outperforms the\nrecent state-of-the-art methods without ground-truth pose annotations and\ndemonstrates substantial generalisation improvements towards unseen scenes.\nRemarkably, our scheme improves the adopted pose estimators substantially even\nwhen initialised with much inferior performance.\n", "date": "2022-12-25", "categories": []}
{"id": "2210.10108", "title": "Parallel Inversion of Neural Radiance Fields for Robust Pose Estimation", "abstract": "  We present a parallelized optimization method based on fast Neural Radiance\nFields (NeRF) for estimating 6-DoF pose of a camera with respect to an object\nor scene. Given a single observed RGB image of the target, we can predict the\ntranslation and rotation of the camera by minimizing the residual between\npixels rendered from a fast NeRF model and pixels in the observed image. We\nintegrate a momentum-based camera extrinsic optimization procedure into Instant\nNeural Graphics Primitives, a recent exceptionally fast NeRF implementation. By\nintroducing parallel Monte Carlo sampling into the pose estimation task, our\nmethod overcomes local minima and improves efficiency in a more extensive\nsearch space. We also show the importance of adopting a more robust pixel-based\nloss function to reduce error. Experiments demonstrate that our method can\nachieve improved generalization and robustness on both synthetic and real-world\nbenchmarks.\n", "date": "2022-10-18", "categories": []}
{"id": "2208.14288", "title": "6IMPOSE: Bridging the Reality Gap in 6D Pose Estimation for Robotic\n  Grasping", "abstract": "  6D pose recognition has been a crucial factor in the success of robotic\ngrasping, and recent deep learning based approaches have achieved remarkable\nresults on benchmarks. However, their generalization capabilities in real-world\napplications remain unclear. To overcome this gap, we introduce 6IMPOSE, a\nnovel framework for sim-to-real data generation and 6D pose estimation. 6IMPOSE\nconsists of four modules: First, a data generation pipeline that employs the 3D\nsoftware suite Blender to create synthetic RGBD image datasets with 6D pose\nannotations. Second, an annotated RGBD dataset of five household objects\ngenerated using the proposed pipeline. Third, a real-time two-stage 6D pose\nestimation approach that integrates the object detector YOLO-V4 and a\nstreamlined, real-time version of the 6D pose estimation algorithm PVN3D\noptimized for time-sensitive robotics applications. Fourth, a codebase designed\nto facilitate the integration of the vision system into a robotic grasping\nexperiment. Our approach demonstrates the efficient generation of large amounts\nof photo-realistic RGBD images and the successful transfer of the trained\ninference model to robotic grasping experiments, achieving an overall success\nrate of 87% in grasping five different household objects from cluttered\nbackgrounds under varying lighting conditions. This is made possible by the\nfine-tuning of data generation and domain randomization techniques, and the\noptimization of the inference pipeline, overcoming the generalization and\nperformance shortcomings of the original PVN3D algorithm. Finally, we make the\ncode, synthetic dataset, and all the pretrained models available on Github.\n", "date": "2022-08-30", "categories": []}
{"id": "2303.05245", "title": "Probabilistic 3d regression with projected huber distribution", "abstract": "  Estimating probability distributions which describe where an object is likely\nto be from camera data is a task with many applications. In this work we\ndescribe properties which we argue such methods should conform to. We also\ndesign a method which conform to these properties. In our experiments we show\nthat our method produces uncertainties which correlate well with empirical\nerrors. We also show that the mode of the predicted distribution outperform our\nregression baselines. The code for our implementation is available online.\n", "date": "2023-03-09", "categories": []}
{"id": "2302.14268", "title": "Self-Supervised Category-Level Articulated Object Pose Estimation with\n  Part-Level SE(3) Equivariance", "abstract": "  Category-level articulated object pose estimation aims to estimate a\nhierarchy of articulation-aware object poses of an unseen articulated object\nfrom a known category. To reduce the heavy annotations needed for supervised\nlearning methods, we present a novel self-supervised strategy that solves this\nproblem without any human labels. Our key idea is to factorize canonical shapes\nand articulated object poses from input articulated shapes through part-level\nequivariant shape analysis. Specifically, we first introduce the concept of\npart-level SE(3) equivariance and devise a network to learn features of such\nproperty. Then, through a carefully designed fine-grained pose-shape\ndisentanglement strategy, we expect that canonical spaces to support pose\nestimation could be induced automatically. Thus, we could further predict\narticulated object poses as per-part rigid transformations describing how parts\ntransform from their canonical part spaces to the camera space. Extensive\nexperiments demonstrate the effectiveness of our method on both complete and\npartial point clouds from synthetic and real articulated object datasets.\n", "date": "2023-02-28", "categories": []}
{"id": "2302.13075", "title": "BOP Challenge 2022 on Detection, Segmentation and Pose Estimation of\n  Specific Rigid Objects", "abstract": "  We present the evaluation methodology, datasets and results of the BOP\nChallenge 2022, the fourth in a series of public competitions organized with\nthe goal to capture the status quo in the field of 6D object pose estimation\nfrom an RGB/RGB-D image. In 2022, we witnessed another significant improvement\nin the pose estimation accuracy -- the state of the art, which was 56.9 AR$_C$\nin 2019 (Vidal et al.) and 69.8 AR$_C$ in 2020 (CosyPose), moved to new heights\nof 83.7 AR$_C$ (GDRNPP). Out of 49 pose estimation methods evaluated since\n2019, the top 18 are from 2022. Methods based on point pair features, which\nwere introduced in 2010 and achieved competitive results even in 2020, are now\nclearly outperformed by deep learning methods. The synthetic-to-real domain gap\nwas again significantly reduced, with 82.7 AR$_C$ achieved by GDRNPP trained\nonly on synthetic images from BlenderProc. The fastest variant of GDRNPP\nreached 80.5 AR$_C$ with an average time per image of 0.23s. Since most of the\nrecent methods for 6D object pose estimation begin by detecting/segmenting\nobjects, we also started evaluating 2D object detection and segmentation\nperformance based on the COCO metrics. Compared to the Mask R-CNN results from\nCosyPose in 2020, detection improved from 60.3 to 77.3 AP$_C$ and segmentation\nfrom 40.5 to 58.7 AP$_C$. The online evaluation system stays open and is\navailable at: \\href{http://bop.felk.cvut.cz/}{bop.felk.cvut.cz}.\n", "date": "2023-02-25", "categories": []}
{"id": "2302.13926", "title": "Image to Sphere: Learning Equivariant Features for Efficient Pose\n  Prediction", "abstract": "  Predicting the pose of objects from a single image is an important but\ndifficult computer vision problem. Methods that predict a single point estimate\ndo not predict the pose of objects with symmetries well and cannot represent\nuncertainty. Alternatively, some works predict a distribution over orientations\nin $\\mathrm{SO}(3)$. However, training such models can be computation- and\nsample-inefficient. Instead, we propose a novel mapping of features from the\nimage domain to the 3D rotation manifold. Our method then leverages\n$\\mathrm{SO}(3)$ equivariant layers, which are more sample efficient, and\noutputs a distribution over rotations that can be sampled at arbitrary\nresolution. We demonstrate the effectiveness of our method at object\norientation prediction, and achieve state-of-the-art performance on the popular\nPASCAL3D+ dataset. Moreover, we show that our method can model complex object\nsymmetries, without any modifications to the parameters or loss function. Code\nis available at https://dmklee.github.io/image2sphere.\n", "date": "2023-02-27", "categories": []}
{"id": "2302.11458", "title": "Fusing Visual Appearance and Geometry for Multi-modality 6DoF Object\n  Tracking", "abstract": "  In many applications of advanced robotic manipulation, six degrees of freedom\n(6DoF) object pose estimates are continuously required. In this work, we\ndevelop a multi-modality tracker that fuses information from visual appearance\nand geometry to estimate object poses. The algorithm extends our previous\nmethod ICG, which uses geometry, to additionally consider surface appearance.\nIn general, object surfaces contain local characteristics from text, graphics,\nand patterns, as well as global differences from distinct materials and colors.\nTo incorporate this visual information, two modalities are developed. For local\ncharacteristics, keypoint features are used to minimize distances between\npoints from keyframes and the current image. For global differences, a novel\nregion approach is developed that considers multiple regions on the object\nsurface. In addition, it allows the modeling of external geometries.\nExperiments on the YCB-Video and OPT datasets demonstrate that our approach\nICG+ performs best on both datasets, outperforming both conventional and deep\nlearning-based methods. At the same time, the algorithm is highly efficient and\nruns at more than 300 Hz. The source code of our tracker is publicly available.\n", "date": "2023-02-22", "categories": []}
{"id": "2302.07360", "title": "Self-supervised learning of object pose estimation using keypoint\n  prediction", "abstract": "  This paper describes recent developments in object specific pose and shape\nprediction from single images. The main contribution is a new approach to\ncamera pose prediction by self-supervised learning of keypoints corresponding\nto locations on a category specific deformable shape. We designed a network to\ngenerate a proxy ground-truth heatmap from a set of keypoints distributed all\nover the category-specific mean shape, where each is represented by a unique\ncolor on a labeled texture. The proxy ground-truth heatmap is used to train a\ndeep keypoint prediction network, which can be used in online inference. The\nproposed approach to camera pose prediction show significant improvements when\ncompared with state-of-the-art methods.\n  Our approach to camera pose prediction is used to infer 3D objects from 2D\nimage frames of video sequences online. To train the reconstruction model, it\nreceives only a silhouette mask from a single frame of a video sequence in\nevery training step and a category-specific mean object shape. We conducted\nexperiments using three different datasets representing the bird category: the\nCUB [51] image dataset, YouTubeVos and the Davis video datasets. The network is\ntrained on the CUB dataset and tested on all three datasets. The online\nexperiments are demonstrated on YouTubeVos and Davis [56] video sequences using\na network trained on the CUB training set.\n", "date": "2023-02-14", "categories": []}
{"id": "2302.07300", "title": "MSDA: Monocular Self-supervised Domain Adaptation for 6D Object Pose\n  Estimation", "abstract": "  Acquiring labeled 6D poses from real images is an expensive and\ntime-consuming task. Though massive amounts of synthetic RGB images are easy to\nobtain, the models trained on them suffer from noticeable performance\ndegradation due to the synthetic-to-real domain gap. To mitigate this\ndegradation, we propose a practical self-supervised domain adaptation approach\nthat takes advantage of real RGB(-D) data without needing real pose labels. We\nfirst pre-train the model with synthetic RGB images and then utilize real\nRGB(-D) images to fine-tune the pre-trained model. The fine-tuning process is\nself-supervised by the RGB-based pose-aware consistency and the depth-guided\nobject distance pseudo-label, which does not require the time-consuming online\ndifferentiable rendering. We build our domain adaptation method based on the\nrecent pose estimator SC6D and evaluate it on the YCB-Video dataset. We\nexperimentally demonstrate that our method achieves comparable performance\nagainst its fully-supervised counterpart while outperforming existing\nstate-of-the-art approaches.\n", "date": "2023-02-14", "categories": []}
{"id": "2210.01112", "title": "Generative Category-Level Shape and Pose Estimation with Semantic\n  Primitives", "abstract": "  Empowering autonomous agents with 3D understanding for daily objects is a\ngrand challenge in robotics applications. When exploring in an unknown\nenvironment, existing methods for object pose estimation are still not\nsatisfactory due to the diversity of object shapes. In this paper, we propose a\nnovel framework for category-level object shape and pose estimation from a\nsingle RGB-D image. To handle the intra-category variation, we adopt a semantic\nprimitive representation that encodes diverse shapes into a unified latent\nspace, which is the key to establish reliable correspondences between observed\npoint clouds and estimated shapes. Then, by using a SIM(3)-invariant shape\ndescriptor, we gracefully decouple the shape and pose of an object, thus\nsupporting latent shape optimization of target objects in arbitrary poses.\nExtensive experiments show that the proposed method achieves SOTA pose\nestimation performance and better generalization in the real-world dataset.\nCode and video are available at https://zju3dv.github.io/gCasp.\n", "date": "2022-10-03", "categories": []}
{"id": "2301.13667", "title": "Collision-aware In-hand 6D Object Pose Estimation using Multiple\n  Vision-based Tactile Sensors", "abstract": "  In this paper, we address the problem of estimating the in-hand 6D pose of an\nobject in contact with multiple vision-based tactile sensors. We reason on the\npossible spatial configurations of the sensors along the object surface.\nSpecifically, we filter contact hypotheses using geometric reasoning and a\nConvolutional Neural Network (CNN), trained on simulated object-agnostic\nimages, to promote those that better comply with the actual tactile images from\nthe sensors. We use the selected sensors configurations to optimize over the\nspace of 6D poses using a Gradient Descent-based approach. We finally rank the\nobtained poses by penalizing those that are in collision with the sensors. We\ncarry out experiments in simulation using the DIGIT vision-based sensor with\nseveral objects, from the standard YCB model set. The results demonstrate that\nour approach estimates object poses that are compatible with actual\nobject-sensor contacts in $87.5\\%$ of cases while reaching an average\npositional error in the order of $2$ centimeters. Our analysis also includes\nqualitative results of experiments with a real DIGIT sensor.\n", "date": "2023-01-31", "categories": []}
{"id": "2212.04632", "title": "Category-Level 6D Object Pose Estimation with Flexible Vector-Based\n  Rotation Representation", "abstract": "  In this paper, we propose a novel 3D graph convolution based pipeline for\ncategory-level 6D pose and size estimation from monocular RGB-D images. The\nproposed method leverages an efficient 3D data augmentation and a novel\nvector-based decoupled rotation representation. Specifically, we first design\nan orientation-aware autoencoder with 3D graph convolution for latent feature\nlearning. The learned latent feature is insensitive to point shift and size\nthanks to the shift and scale-invariance properties of the 3D graph\nconvolution. Then, to efficiently decode the rotation information from the\nlatent feature, we design a novel flexible vector-based decomposable rotation\nrepresentation that employs two decoders to complementarily access the rotation\ninformation. The proposed rotation representation has two major advantages: 1)\ndecoupled characteristic that makes the rotation estimation easier; 2) flexible\nlength and rotated angle of the vectors allow us to find a more suitable vector\nrepresentation for specific pose estimation task. Finally, we propose a 3D\ndeformation mechanism to increase the generalization ability of the pipeline.\nExtensive experiments show that the proposed pipeline achieves state-of-the-art\nperformance on category-level tasks. Further, the experiments demonstrate that\nthe proposed rotation representation is more suitable for the pose estimation\ntasks than other rotation representations.\n", "date": "2022-12-09", "categories": []}
{"id": "2301.08147", "title": "RGB-D-Based Categorical Object Pose and Shape Estimation: Methods,\n  Datasets, and Evaluation", "abstract": "  Recently, various methods for 6D pose and shape estimation of objects at a\nper-category level have been proposed. This work provides an overview of the\nfield in terms of methods, datasets, and evaluation protocols. First, an\noverview of existing works and their commonalities and differences is provided.\nSecond, we take a critical look at the predominant evaluation protocol,\nincluding metrics and datasets. Based on the findings, we propose a new set of\nmetrics, contribute new annotations for the Redwood dataset, and evaluate\nstate-of-the-art methods in a fair comparison. The results indicate that\nexisting methods do not generalize well to unconstrained orientations and are\nactually heavily biased towards objects being upright. We provide an\neasy-to-use evaluation toolbox with well-defined metrics, methods, and dataset\ninterfaces, which allows evaluation and comparison with various\nstate-of-the-art approaches\n(https://github.com/roym899/pose_and_shape_evaluation).\n", "date": "2023-01-19", "categories": []}
{"id": "2301.07673", "title": "OnePose++: Keypoint-Free One-Shot Object Pose Estimation without CAD\n  Models", "abstract": "  We propose a new method for object pose estimation without CAD models. The\nprevious feature-matching-based method OnePose has shown promising results\nunder a one-shot setting which eliminates the need for CAD models or\nobject-specific training. However, OnePose relies on detecting repeatable image\nkeypoints and is thus prone to failure on low-textured objects. We propose a\nkeypoint-free pose estimation pipeline to remove the need for repeatable\nkeypoint detection. Built upon the detector-free feature matching method LoFTR,\nwe devise a new keypoint-free SfM method to reconstruct a semi-dense\npoint-cloud model for the object. Given a query image for object pose\nestimation, a 2D-3D matching network directly establishes 2D-3D correspondences\nbetween the query image and the reconstructed point-cloud model without first\ndetecting keypoints in the image. Experiments show that the proposed pipeline\noutperforms existing one-shot CAD-model-free methods by a large margin and is\ncomparable to CAD-model-based methods on LINEMOD even for low-textured objects.\nWe also collect a new dataset composed of 80 sequences of 40 low-textured\nobjects to facilitate future research on one-shot object pose estimation. The\nsupplementary material, code and dataset are available on the project page:\nhttps://zju3dv.github.io/onepose_plus_plus/.\n", "date": "2023-01-18", "categories": []}
{"id": "2212.13415", "title": "Spacecraft Pose Estimation Based on Unsupervised Domain Adaptation and\n  on a 3D-Guided Loss Combination", "abstract": "  Spacecraft pose estimation is a key task to enable space missions in which\ntwo spacecrafts must navigate around each other. Current state-of-the-art\nalgorithms for pose estimation employ data-driven techniques. However, there is\nan absence of real training data for spacecraft imaged in space conditions due\nto the costs and difficulties associated with the space environment. This has\nmotivated the introduction of 3D data simulators, solving the issue of data\navailability but introducing a large gap between the training (source) and test\n(target) domains. We explore a method that incorporates 3D structure into the\nspacecraft pose estimation pipeline to provide robustness to intensity domain\nshift and we present an algorithm for unsupervised domain adaptation with\nrobust pseudo-labelling. Our solution has ranked second in the two categories\nof the 2021 Pose Estimation Challenge organised by the European Space Agency\nand the Stanford University, achieving the lowest average error over the two\ncategories.\n", "date": "2022-12-27", "categories": []}
{"id": "2212.12668", "title": "Differentiable Rendering for Pose Estimation in Proximity Operations", "abstract": "  Differentiable rendering aims to compute the derivative of the image\nrendering function with respect to the rendering parameters. This paper\npresents a novel algorithm for 6-DoF pose estimation through gradient-based\noptimization using a differentiable rendering pipeline. We emphasize two key\ncontributions: (1) instead of solving the conventional 2D to 3D correspondence\nproblem and computing reprojection errors, images (rendered using the 3D model)\nare compared only in the 2D feature space via sparse 2D feature\ncorrespondences. (2) Instead of an analytical image formation model, we compute\nan approximate local gradient of the rendering process through online learning.\nThe learning data consists of image features extracted from multi-viewpoint\nrenders at small perturbations in the pose neighborhood. The gradients are\npropagated through the rendering pipeline for the 6-DoF pose estimation using\nnonlinear least squares. This gradient-based optimization regresses directly\nupon the pose parameters by aligning the 3D model to reproduce a reference\nimage shape. Using representative experiments, we demonstrate the application\nof our approach to pose estimation in proximity operations.\n", "date": "2022-12-24", "categories": []}
{"id": "2212.12103", "title": "Bridging the Domain Gap in Satellite Pose Estimation: a Self-Training\n  Approach based on Geometrical Constraints", "abstract": "  Recently, unsupervised domain adaptation in satellite pose estimation has\ngained increasing attention, aiming at alleviating the annotation cost for\ntraining deep models. To this end, we propose a self-training framework based\non the domain-agnostic geometrical constraints. Specifically, we train a neural\nnetwork to predict the 2D keypoints of a satellite and then use PnP to estimate\nthe pose. The poses of target samples are regarded as latent variables to\nformulate the task as a minimization problem. Furthermore, we leverage\nfine-grained segmentation to tackle the information loss issue caused by\nabstracting the satellite as sparse keypoints. Finally, we iteratively solve\nthe minimization problem in two steps: pseudo-label generation and network\ntraining. Experimental results show that our method adapts well to the target\ndomain. Moreover, our method won the 1st place on the sunlamp task of the\nsecond international Satellite Pose Estimation Competition.\n", "date": "2022-12-23", "categories": []}
{"id": "2212.06870", "title": "MegaPose: 6D Pose Estimation of Novel Objects via Render & Compare", "abstract": "  We introduce MegaPose, a method to estimate the 6D pose of novel objects,\nthat is, objects unseen during training. At inference time, the method only\nassumes knowledge of (i) a region of interest displaying the object in the\nimage and (ii) a CAD model of the observed object. The contributions of this\nwork are threefold. First, we present a 6D pose refiner based on a\nrender&compare strategy which can be applied to novel objects. The shape and\ncoordinate system of the novel object are provided as inputs to the network by\nrendering multiple synthetic views of the object's CAD model. Second, we\nintroduce a novel approach for coarse pose estimation which leverages a network\ntrained to classify whether the pose error between a synthetic rendering and an\nobserved image of the same object can be corrected by the refiner. Third, we\nintroduce a large-scale synthetic dataset of photorealistic images of thousands\nof objects with diverse visual and shape properties and show that this\ndiversity is crucial to obtain good generalization performance on novel\nobjects. We train our approach on this large synthetic dataset and apply it\nwithout retraining to hundreds of novel objects in real images from several\npose estimation benchmarks. Our approach achieves state-of-the-art performance\non the ModelNet and YCB-Video datasets. An extensive evaluation on the 7 core\ndatasets of the BOP challenge demonstrates that our approach achieves\nperformance competitive with existing approaches that require access to the\ntarget objects during training. Code, dataset and trained models are available\non the project page: https://megapose6d.github.io/.\n", "date": "2022-12-13", "categories": []}
{"id": "2212.06300", "title": "Accidental Turntables: Learning 3D Pose by Watching Objects Turn", "abstract": "  We propose a technique for learning single-view 3D object pose estimation\nmodels by utilizing a new source of data -- in-the-wild videos where objects\nturn. Such videos are prevalent in practice (e.g., cars in roundabouts,\nairplanes near runways) and easy to collect. We show that classical\nstructure-from-motion algorithms, coupled with the recent advances in instance\ndetection and feature matching, provides surprisingly accurate relative 3D pose\nestimation on such videos. We propose a multi-stage training scheme that first\nlearns a canonical pose across a collection of videos and then supervises a\nmodel for single-view pose estimation. The proposed technique achieves\ncompetitive performance with respect to existing state-of-the-art on standard\nbenchmarks for 3D pose estimation, without requiring any pose labels during\ntraining. We also contribute an Accidental Turntables Dataset, containing a\nchallenging set of 41,212 images of cars in cluttered backgrounds, motion blur\nand illumination changes that serves as a benchmark for 3D pose estimation.\n", "date": "2022-12-13", "categories": []}
{"id": "2212.05560", "title": "Context-aware 6D Pose Estimation of Known Objects using RGB-D data", "abstract": "  6D object pose estimation has been a research topic in the field of computer\nvision and robotics. Many modern world applications like robot grasping,\nmanipulation, autonomous navigation etc, require the correct pose of objects\npresent in a scene to perform their specific task. It becomes even harder when\nthe objects are placed in a cluttered scene and the level of occlusion is high.\nPrior works have tried to overcome this problem but could not achieve accuracy\nthat can be considered reliable in real-world applications. In this paper, we\npresent an architecture that, unlike prior work, is context-aware. It utilizes\nthe context information available to us about the objects. Our proposed\narchitecture treats the objects separately according to their types i.e;\nsymmetric and non-symmetric. A deeper estimator and refiner network pair is\nused for non-symmetric objects as compared to symmetric due to their intrinsic\ndifferences. Our experiments show an enhancement in the accuracy of about 3.2%\nover the LineMOD dataset, which is considered a benchmark for pose estimation\nin the occluded and cluttered scenes, against the prior state-of-the-art\nDenseFusion. Our results also show that the inference time we got is sufficient\nfor real-time usage.\n", "date": "2022-12-11", "categories": []}
{"id": "2210.05318", "title": "CASAPose: Class-Adaptive and Semantic-Aware Multi-Object Pose Estimation", "abstract": "  Applications in the field of augmented reality or robotics often require\njoint localisation and 6D pose estimation of multiple objects. However, most\nalgorithms need one network per object class to be trained in order to provide\nthe best results. Analysing all visible objects demands multiple inferences,\nwhich is memory and time-consuming. We present a new single-stage architecture\ncalled CASAPose that determines 2D-3D correspondences for pose estimation of\nmultiple different objects in RGB images in one pass. It is fast and memory\nefficient, and achieves high accuracy for multiple objects by exploiting the\noutput of a semantic segmentation decoder as control input to a keypoint\nrecognition decoder via local class-adaptive normalisation. Our new\ndifferentiable regression of keypoint locations significantly contributes to a\nfaster closing of the domain gap between real test and synthetic training data.\nWe apply segmentation-aware convolutions and upsampling operations to increase\nthe focus inside the object mask and to reduce mutual interference of occluding\nobjects. For each inserted object, the network grows by only one output\nsegmentation map and a negligible number of parameters. We outperform\nstate-of-the-art approaches in challenging multi-object scenes with\ninter-object occlusion and synthetic training.\n", "date": "2022-10-11", "categories": []}
{"id": "2211.14125", "title": "PoET: Pose Estimation Transformer for Single-View, Multi-Object 6D Pose\n  Estimation", "abstract": "  Accurate 6D object pose estimation is an important task for a variety of\nrobotic applications such as grasping or localization. It is a challenging task\ndue to object symmetries, clutter and occlusion, but it becomes more\nchallenging when additional information, such as depth and 3D models, is not\nprovided. We present a transformer-based approach that takes an RGB image as\ninput and predicts a 6D pose for each object in the image. Besides the image,\nour network does not require any additional information such as depth maps or\n3D object models. First, the image is passed through an object detector to\ngenerate feature maps and to detect objects. Then, the feature maps are fed\ninto a transformer with the detected bounding boxes as additional information.\nAfterwards, the output object queries are processed by a separate translation\nand rotation head. We achieve state-of-the-art results for RGB-only approaches\non the challenging YCB-V dataset. We illustrate the suitability of the\nresulting model as pose sensor for a 6-DoF state estimation task. Code is\navailable at https://github.com/aau-cns/poet.\n", "date": "2022-11-25", "categories": []}
{"id": "2211.10470", "title": "A mixed-reality dataset for category-level 6D pose and size estimation\n  of hand-occluded containers", "abstract": "  Estimating the 6D pose and size of household containers is challenging due to\nlarge intra-class variations in the object properties, such as shape, size,\nappearance, and transparency. The task is made more difficult when these\nobjects are held and manipulated by a person due to varying degrees of hand\nocclusions caused by the type of grasps and by the viewpoint of the camera\nobserving the person holding the object. In this paper, we present a\nmixed-reality dataset of hand-occluded containers for category-level 6D object\npose and size estimation. The dataset consists of 138,240 images of rendered\nhands and forearms holding 48 synthetic objects, split into 3 grasp categories\nover 30 real backgrounds. We re-train and test an existing model for 6D object\npose estimation on our mixed-reality dataset. We discuss the impact of the use\nof this dataset in improving the task of 6D pose and size estimation.\n", "date": "2022-11-18", "categories": []}
{"id": "2211.11394", "title": "Learning Implicit Probability Distribution Functions for Symmetric\n  Orientation Estimation from RGB Images Without Pose Labels", "abstract": "  Object pose estimation is a necessary prerequisite for autonomous robotic\nmanipulation, but the presence of symmetry increases the complexity of the pose\nestimation task. Existing methods for object pose estimation output a single 6D\npose. Thus, they lack the ability to reason about symmetries. Lately, modeling\nobject orientation as a non-parametric probability distribution on the SO(3)\nmanifold by neural networks has shown impressive results. However, acquiring\nlarge-scale datasets to train pose estimation models remains a bottleneck. To\naddress this limitation, we introduce an automatic pose labeling scheme. Given\nRGB-D images without object pose annotations and 3D object models, we design a\ntwo-stage pipeline consisting of point cloud registration and\nrender-and-compare validation to generate multiple symmetrical\npseudo-ground-truth pose labels for each image. Using the generated pose\nlabels, we train an ImplicitPDF model to estimate the likelihood of an\norientation hypothesis given an RGB image. An efficient hierarchical sampling\nof the SO(3) manifold enables tractable generation of the complete set of\nsymmetries at multiple resolutions. During inference, the most likely\norientation of the target object is estimated using gradient ascent. We\nevaluate the proposed automatic pose labeling scheme and the ImplicitPDF model\non a photorealistic dataset and the T-Less dataset, demonstrating the\nadvantages of the proposed method.\n", "date": "2022-11-21", "categories": []}
{"id": "2207.08925", "title": "Image to Icosahedral Projection for $\\mathrm{SO}(3)$ Object Reasoning\n  from Single-View Images", "abstract": "  Reasoning about 3D objects based on 2D images is challenging due to\nvariations in appearance caused by viewing the object from different\norientations. Tasks such as object classification are invariant to 3D rotations\nand other such as pose estimation are equivariant. However, imposing\nequivariance as a model constraint is typically not possible with 2D image\ninput because we do not have an a priori model of how the image changes under\nout-of-plane object rotations. The only $\\mathrm{SO}(3)$-equivariant models\nthat currently exist require point cloud or voxel input rather than 2D images.\nIn this paper, we propose a novel architecture based on icosahedral group\nconvolutions that reasons in $\\mathrm{SO(3)}$ by learning a projection of the\ninput image onto an icosahedron. The resulting model is approximately\nequivariant to rotation in $\\mathrm{SO}(3)$. We apply this model to object pose\nestimation and shape classification tasks and find that it outperforms\nreasonable baselines. Project website:\n\\url{https://dmklee.github.io/image2icosahedral}\n", "date": "2022-07-18", "categories": []}
{"id": "2211.07491", "title": "Piecewise Planar Hulls for Semi-Supervised Learning of 3D Shape and Pose\n  from 2D Images", "abstract": "  We study the problem of estimating 3D shape and pose of an object in terms of\nkeypoints, from a single 2D image.\n  The shape and pose are learned directly from images collected by categories\nand their partial 2D keypoint annotations.. In this work, we first propose an\nend-to-end training framework for intermediate 2D keypoints extraction and\nfinal 3D shape and pose estimation. The proposed framework is then trained\nusing only the weak supervision of the intermediate 2D keypoints. Additionally,\nwe devise a semi-supervised training framework that benefits from both labeled\nand unlabeled data. To leverage the unlabeled data, we introduce and exploit\nthe \\emph{piece-wise planar hull} prior of the canonical object shape. These\nplanar hulls are defined manually once per object category, with the help of\nthe keypoints. On the one hand, the proposed method learns to segment these\nplanar hulls from the labeled data. On the other hand, it simultaneously\nenforces the consistency between predicted keypoints and the segmented hulls on\nthe unlabeled data. The enforced consistency allows us to efficiently use the\nunlabeled data for the task at hand. The proposed method achieves comparable\nresults with fully supervised state-of-the-art methods by using only half of\nthe annotations. Our source code will be made publicly available.\n", "date": "2022-11-14", "categories": []}
{"id": "2211.04691", "title": "A Solution for a Fundamental Problem of 3D Inference based on 2D\n  Representations", "abstract": "  3D inference from monocular vision using neural networks is an important\nresearch area of computer vision. Applications of the research area are various\nwith many proposed solutions and have shown remarkable performance. Although\nmany efforts have been invested, there are still unanswered questions, some of\nwhich are fundamental. In this paper, I discuss a problem that I hope will come\nto be known as a generalization of the Blind Perspective-n-Point (Blind PnP)\nproblem for object-driven 3D inference based on 2D representations. The vital\ndifference between the fundamental problem and the Blind PnP problem is that 3D\ninference parameters in the fundamental problem are attached directly to 3D\npoints and the camera concept will be represented through the sharing of the\nparameters of these points. By providing an explainable and robust\ngradient-decent solution based on 2D representations for an important special\ncase of the problem, the paper opens up a new approach for using available\ninformation-based learning methods to solve problems related to 3D object pose\nestimation from 2D images.\n", "date": "2022-11-09", "categories": []}
{"id": "2211.03211", "title": "Towards real-time 6D pose estimation of objects in single-view cone-beam\n  X-ray", "abstract": "  Deep learning-based pose estimation algorithms can successfully estimate the\npose of objects in an image, especially in the field of color images. 6D Object\npose estimation based on deep learning models for X-ray images often use custom\narchitectures that employ extensive CAD models and simulated data for training\npurposes. Recent RGB-based methods opt to solve pose estimation problems using\nsmall datasets, making them more attractive for the X-ray domain where medical\ndata is scarcely available. We refine an existing RGB-based model\n(SingleShotPose) to estimate the 6D pose of a marked cube from grayscale X-ray\nimages by creating a generic solution trained on only real X-ray data and\nadjusted for X-ray acquisition geometry. The model regresses 2D control points\nand calculates the pose through 2D/3D correspondences using\nPerspective-n-Point(PnP), allowing a single trained model to be used across all\nsupporting cone-beam-based X-ray geometries. Since modern X-ray systems\ncontinuously adjust acquisition parameters during a procedure, it is essential\nfor such a pose estimation network to consider these parameters in order to be\ndeployed successfully and find a real use case. With a 5-cm/5-degree accuracy\nof 93% and an average 3D rotation error of 2.2 degrees, the results of the\nproposed approach are comparable with state-of-the-art alternatives, while\nrequiring significantly less real training examples and being applicable in\nreal-time applications.\n", "date": "2022-11-06", "categories": []}
{"id": "2211.01644", "title": "StereoPose: Category-Level 6D Transparent Object Pose Estimation from\n  Stereo Images via Back-View NOCS", "abstract": "  Most existing methods for category-level pose estimation rely on object point\nclouds. However, when considering transparent objects, depth cameras are\nusually not able to capture meaningful data, resulting in point clouds with\nsevere artifacts. Without a high-quality point cloud, existing methods are not\napplicable to challenging transparent objects. To tackle this problem, we\npresent StereoPose, a novel stereo image framework for category-level object\npose estimation, ideally suited for transparent objects. For a robust\nestimation from pure stereo images, we develop a pipeline that decouples\ncategory-level pose estimation into object size estimation, initial pose\nestimation, and pose refinement. StereoPose then estimates object pose based on\nrepresentation in the normalized object coordinate space~(NOCS). To address the\nissue of image content aliasing, we further define a back-view NOCS map for the\ntransparent object. The back-view NOCS aims to reduce the network learning\nambiguity caused by content aliasing, and leverage informative cues on the back\nof the transparent object for more accurate pose estimation. To further improve\nthe performance of the stereo framework, StereoPose is equipped with a parallax\nattention module for stereo feature fusion and an epipolar loss for improving\nthe stereo-view consistency of network predictions. Extensive experiments on\nthe public TOD dataset demonstrate the superiority of the proposed StereoPose\nframework for category-level 6D transparent object pose estimation.\n", "date": "2022-11-03", "categories": []}
{"id": "2211.00960", "title": "Ambiguity-Aware Multi-Object Pose Optimization for Visually-Assisted\n  Robot Manipulation", "abstract": "  6D object pose estimation aims to infer the relative pose between the object\nand the camera using a single image or multiple images. Most works have focused\non predicting the object pose without associated uncertainty under occlusion\nand structural ambiguity (symmetricity). However, these works demand prior\ninformation about shape attributes, and this condition is hardly satisfied in\nreality; even asymmetric objects may be symmetric under the viewpoint change.\nIn addition, acquiring and fusing diverse sensor data is challenging when\nextending them to robotics applications. Tackling these limitations, we present\nan ambiguity-aware 6D object pose estimation network, PrimA6D++, as a generic\nuncertainty prediction method. The major challenges in pose estimation, such as\nocclusion and symmetry, can be handled in a generic manner based on the\nmeasured ambiguity of the prediction. Specifically, we devise a network to\nreconstruct the three rotation axis primitive images of a target object and\npredict the underlying uncertainty along each primitive axis. Leveraging the\nestimated uncertainty, we then optimize multi-object poses using visual\nmeasurements and camera poses by treating it as an object SLAM problem. The\nproposed method shows a significant performance improvement in T-LESS and\nYCB-Video datasets. We further demonstrate real-time scene recognition\ncapability for visually-assisted robot manipulation. Our code and supplementary\nmaterials are available at https://github.com/rpmsnu/PrimA6D.\n", "date": "2022-11-02", "categories": []}
{"id": "2210.11718", "title": "CRT-6D: Fast 6D Object Pose Estimation with Cascaded Refinement\n  Transformers", "abstract": "  Learning based 6D object pose estimation methods rely on computing large\nintermediate pose representations and/or iteratively refining an initial\nestimation with a slow render-compare pipeline. This paper introduces a novel\nmethod we call Cascaded Pose Refinement Transformers, or CRT-6D. We replace the\ncommonly used dense intermediate representation with a sparse set of features\nsampled from the feature pyramid we call OSKFs(Object Surface Keypoint\nFeatures) where each element corresponds to an object keypoint. We employ\nlightweight deformable transformers and chain them together to iteratively\nrefine proposed poses over the sampled OSKFs. We achieve inference runtimes 2x\nfaster than the closest real-time state of the art methods while supporting up\nto 21 objects on a single model. We demonstrate the effectiveness of CRT-6D by\nperforming extensive experiments on the LM-O and YCBV datasets. Compared to\nreal-time methods, we achieve state of the art on LM-O and YCB-V, falling\nslightly behind methods with inference runtimes one order of magnitude higher.\nThe source code is available at: https://github.com/PedroCastro/CRT-6D\n", "date": "2022-10-21", "categories": []}
{"id": "2210.11973", "title": "Real-Time Constrained 6D Object-Pose Tracking of An In-Hand Suture\n  Needle for Minimally Invasive Robotic Surgery", "abstract": "  Autonomous suturing has been a long-sought-after goal for surgical robotics.\nOutside of staged environments, accurate localization of suture needles is a\ncritical foundation for automating various suture needle manipulation tasks in\nthe real world. When localizing a needle held by a gripper, previous work\nusually tracks them separately without considering their relationship. Because\nof the significant errors that can arise in the stereo-triangulation of objects\nand instruments, their reconstructions may often not be consistent. This can\nlead to unrealistic tool-needle grasp reconstructions that are infeasible.\nInstead, an obvious strategy to improve localization would be to leverage\nconstraints that arise from contact, thereby constraining reconstructions of\nobjects and instruments into a jointly feasible space. In this work, we\nconsider feasible grasping constraints when tracking the 6D pose of an in-hand\nsuture needle. We propose a reparameterization trick to define a new state\nspace for describing a needle pose, where grasp constraints can be easily\ndefined and satisfied. Our proposed state space and feasible grasping\nconstraints are then incorporated into Bayesian filters for real-time needle\nlocalization. In the experiments, we show that our constrained methods\noutperform previous unconstrained/constrained tracking approaches and\ndemonstrate the importance of incorporating feasible grasping constraints into\nautomating suture needle manipulation tasks.\n", "date": "2022-10-21", "categories": []}
{"id": "2210.08123", "title": "Keypoint Cascade Voting for Point Cloud Based 6DoF Pose Estimation", "abstract": "  We propose a novel keypoint voting 6DoF object pose estimation method, which\ntakes pure unordered point cloud geometry as input without RGB information. The\nproposed cascaded keypoint voting method, called RCVPose3D, is based upon a\nnovel architecture which separates the task of semantic segmentation from that\nof keypoint regression, thereby increasing the effectiveness of both and\nimproving the ultimate performance. The method also introduces a pairwise\nconstraint in between different keypoints to the loss function when regressing\nthe quantity for keypoint estimation, which is shown to be effective, as well\nas a novel Voter Confident Score which enhances both the learning and inference\nstages. Our proposed RCVPose3D achieves state-of-the-art performance on the\nOcclusion LINEMOD (74.5%) and YCB-Video (96.9%) datasets, outperforming\nexisting pure RGB and RGB-D based methods, as well as being competitive with\nRGB plus point cloud methods.\n", "date": "2022-10-14", "categories": []}
{"id": "2208.06195", "title": "Category-Level Pose Retrieval with Contrastive Features Learnt with\n  Occlusion Augmentation", "abstract": "  Pose estimation is usually tackled as either a bin classification or a\nregression problem. In both cases, the idea is to directly predict the pose of\nan object. This is a non-trivial task due to appearance variations between\nsimilar poses and similarities between dissimilar poses. Instead, we follow the\nkey idea that comparing two poses is easier than directly predicting one.\nRender-and-compare approaches have been employed to that end, however, they\ntend to be unstable, computationally expensive, and slow for real-time\napplications. We propose doing category-level pose estimation by learning an\nalignment metric in an embedding space using a contrastive loss with a dynamic\nmargin and a continuous pose-label space. For efficient inference, we use a\nsimple real-time image retrieval scheme with a pre-rendered and pre-embedded\nreference set of renderings. To achieve robustness to real-world conditions, we\nemploy synthetic occlusions, bounding box perturbations, and appearance\naugmentations. Our approach achieves state-of-the-art performance on PASCAL3D\nand OccludedPASCAL3D and surpasses the competing methods on KITTI3D in a\ncross-dataset evaluation setting. The code is currently available at\nhttps://github.com/gkouros/contrastive-pose-retrieval.\n", "date": "2022-08-12", "categories": []}
{"id": "2210.05232", "title": "DCL-Net: Deep Correspondence Learning Network for 6D Pose Estimation", "abstract": "  Establishment of point correspondence between camera and object coordinate\nsystems is a promising way to solve 6D object poses. However, surrogate\nobjectives of correspondence learning in 3D space are a step away from the true\nones of object pose estimation, making the learning suboptimal for the end\ntask. In this paper, we address this shortcoming by introducing a new method of\nDeep Correspondence Learning Network for direct 6D object pose estimation,\nshortened as DCL-Net. Specifically, DCL-Net employs dual newly proposed Feature\nDisengagement and Alignment (FDA) modules to establish, in the feature space,\npartial-to-partial correspondence and complete-to-complete one for partial\nobject observation and its complete CAD model, respectively, which result in\naggregated pose and match feature pairs from two coordinate systems; these two\nFDA modules thus bring complementary advantages. The match feature pairs are\nused to learn confidence scores for measuring the qualities of deep\ncorrespondence, while the pose feature pairs are weighted by confidence scores\nfor direct object pose regression. A confidence-based pose refinement network\nis also proposed to further improve pose precision in an iterative manner.\nExtensive experiments show that DCL-Net outperforms existing methods on three\nbenchmarking datasets, including YCB-Video, LineMOD, and Oclussion-LineMOD;\nablation studies also confirm the efficacy of our novel designs.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.06001", "title": "Estimating the Pose of a Euro Pallet with an RGB Camera based on\n  Synthetic Training Data", "abstract": "  Estimating the pose of a pallet and other logistics objects is crucial for\nvarious use cases, such as automatized material handling or tracking.\nInnovations in computer vision, computing power, and machine learning open up\nnew opportunities for device-free localization based on cameras and neural\nnetworks. Large image datasets with annotated poses are required for training\nthe network. Manual annotation, especially of 6D poses, is an extremely\nlabor-intensive process. Hence, newer approaches often leverage synthetic\ntraining data to automatize the process of generating annotated image datasets.\nIn this work, the generation of synthetic training data for 6D pose estimation\nof pallets is presented. The data is then used to train the Deep Object Pose\nEstimation (DOPE) algorithm. The experimental validation of the algorithm\nproves that the 6D pose estimation of a standardized Euro pallet with a\nRed-Green-Blue (RGB) camera is feasible. The comparison of the results from\nthree varying datasets under different lighting conditions shows the relevance\nof an appropriate dataset design to achieve an accurate and robust\nlocalization. The quantitative evaluation shows an average position error of\nless than 20 cm for the preferred dataset. The validated training dataset and a\nphotorealistic model of a Euro pallet are publicly provided.\n", "date": "2022-10-12", "categories": []}
{"id": "2209.07589", "title": "PIZZA: A Powerful Image-only Zero-Shot Zero-CAD Approach to 6 DoF\n  Tracking", "abstract": "  Estimating the relative pose of a new object without prior knowledge is a\nhard problem, while it is an ability very much needed in robotics and Augmented\nReality. We present a method for tracking the 6D motion of objects in RGB video\nsequences when neither the training images nor the 3D geometry of the objects\nare available. In contrast to previous works, our method can therefore consider\nunknown objects in open world instantly, without requiring any prior\ninformation or a specific training phase. We consider two architectures, one\nbased on two frames, and the other relying on a Transformer Encoder, which can\nexploit an arbitrary number of past frames. We train our architectures using\nonly synthetic renderings with domain randomization. Our results on challenging\ndatasets are on par with previous works that require much more information\n(training images of the target objects, 3D models, and/or depth data). Our\nsource code is available at https://github.com/nv-nguyen/pizza\n", "date": "2022-09-15", "categories": []}
{"id": "2210.01044", "title": "SPARC: Sparse Render-and-Compare for CAD model alignment in a single RGB\n  image", "abstract": "  Estimating 3D shapes and poses of static objects from a single image has\nimportant applications for robotics, augmented reality and digital content\ncreation. Often this is done through direct mesh predictions which produces\nunrealistic, overly tessellated shapes or by formulating shape prediction as a\nretrieval task followed by CAD model alignment. Directly predicting CAD model\nposes from 2D image features is difficult and inaccurate. Some works, such as\nROCA, regress normalised object coordinates and use those for computing poses.\nWhile this can produce more accurate pose estimates, predicting normalised\nobject coordinates is susceptible to systematic failure. Leveraging efficient\ntransformer architectures we demonstrate that a sparse, iterative,\nrender-and-compare approach is more accurate and robust than relying on\nnormalised object coordinates. For this we combine 2D image information\nincluding sparse depth and surface normal values which we estimate directly\nfrom the image with 3D CAD model information in early fusion. In particular, we\nreproject points sampled from the CAD model in an initial, random pose and\ncompute their depth and surface normal values. This combined information is the\ninput to a pose prediction network, SPARC-Net which we train to predict a 9 DoF\nCAD model pose update. The CAD model is reprojected again and the next pose\nupdate is predicted. Our alignment procedure converges after just 3 iterations,\nimproving the state-of-the-art performance on the challenging real-world\ndataset ScanNet from 25.0% to 31.8% instance alignment accuracy. Code will be\nreleased at https://github.com/florianlanger/SPARC .\n", "date": "2022-10-03", "categories": []}
{"id": "2208.00237", "title": "RBP-Pose: Residual Bounding Box Projection for Category-Level Pose\n  Estimation", "abstract": "  Category-level object pose estimation aims to predict the 6D pose as well as\nthe 3D metric size of arbitrary objects from a known set of categories. Recent\nmethods harness shape prior adaptation to map the observed point cloud into the\ncanonical space and apply Umeyama algorithm to recover the pose and size.\nHowever, their shape prior integration strategy boosts pose estimation\nindirectly, which leads to insufficient pose-sensitive feature extraction and\nslow inference speed. To tackle this problem, in this paper, we propose a novel\ngeometry-guided Residual Object Bounding Box Projection network RBP-Pose that\njointly predicts object pose and residual vectors describing the displacements\nfrom the shape-prior-indicated object surface projections on the bounding box\ntowards the real surface projections. Such definition of residual vectors is\ninherently zero-mean and relatively small, and explicitly encapsulates spatial\ncues of the 3D object for robust and accurate pose regression. We enforce\ngeometry-aware consistency terms to align the predicted pose and residual\nvectors to further boost performance.\n", "date": "2022-07-30", "categories": []}
{"id": "2208.02129", "title": "SC6D: Symmetry-agnostic and Correspondence-free 6D Object Pose\n  Estimation", "abstract": "  This paper presents an efficient symmetry-agnostic and correspondence-free\nframework, referred to as SC6D, for 6D object pose estimation from a single\nmonocular RGB image. SC6D requires neither the 3D CAD model of the object nor\nany prior knowledge of the symmetries. The pose estimation is decomposed into\nthree sub-tasks: a) object 3D rotation representation learning and matching; b)\nestimation of the 2D location of the object center; and c) scale-invariant\ndistance estimation (the translation along the z-axis) via classification. SC6D\nis evaluated on three benchmark datasets, T-LESS, YCB-V, and ITODD, and results\nin state-of-the-art performance on the T-LESS dataset. Moreover, SC6D is\ncomputationally much more efficient than the previous state-of-the-art method\nSurfEmb. The implementation and pre-trained models are publicly available at\nhttps://github.com/dingdingcai/SC6D-pose.\n", "date": "2022-08-03", "categories": []}
{"id": "2209.08266", "title": "6DOF Pose Estimation of a 3D Rigid Object based on Edge-enhanced Point\n  Pair Features", "abstract": "  The point pair feature (PPF) is widely used for 6D pose estimation. In this\npaper, we propose an efficient 6D pose estimation method based on the PPF\nframework. We introduce a well-targeted down-sampling strategy that focuses\nmore on edge area for efficient feature extraction of complex geometry. A pose\nhypothesis validation approach is proposed to resolve the symmetric ambiguity\nby calculating edge matching degree. We perform evaluations on two challenging\ndatasets and one real-world collected dataset, demonstrating the superiority of\nour method on pose estimation of geometrically complex, occluded, symmetrical\nobjects. We further validate our method by applying it to simulated punctures.\n", "date": "2022-09-17", "categories": []}
{"id": "2209.06562", "title": "INV-Flow2PoseNet: Light-Resistant Rigid Object Pose from Optical Flow of\n  RGB-D Images using Images, Normals and Vertices", "abstract": "  This paper presents a novel architecture for simultaneous estimation of\nhighly accurate optical flows and rigid scene transformations for difficult\nscenarios where the brightness assumption is violated by strong shading\nchanges. In the case of rotating objects or moving light sources, such as those\nencountered for driving cars in the dark, the scene appearance often changes\nsignificantly from one view to the next. Unfortunately, standard methods for\ncalculating optical flows or poses are based on the expectation that the\nappearance of features in the scene remain constant between views. These\nmethods may fail frequently in the investigated cases. The presented method\nfuses texture and geometry information by combining image, vertex and normal\ndata to compute an illumination-invariant optical flow. By using a\ncoarse-to-fine strategy, globally anchored optical flows are learned, reducing\nthe impact of erroneous shading-based pseudo-correspondences. Based on the\nlearned optical flows, a second architecture is proposed that predicts robust\nrigid transformations from the warped vertex and normal maps. Particular\nattention is payed to situations with strong rotations, which often cause such\nshading changes. Therefore a 3-step procedure is proposed that profitably\nexploits correlations between the normals and vertices. The method has been\nevaluated on a newly created dataset containing both synthetic and real data\nwith strong rotations and shading effects. This data represents the typical use\ncase in 3D reconstruction, where the object often rotates in large steps\nbetween the partial reconstructions. Additionally, we apply the method to the\nwell-known Kitti Odometry dataset. Even if, due to fulfillment of the brighness\nassumption, this is not the typical use case of the method, the applicability\nto standard situations and the relation to other methods is therefore\nestablished.\n", "date": "2022-09-14", "categories": []}
{"id": "2209.05624", "title": "Robust Category-Level 6D Pose Estimation with Coarse-to-Fine Rendering\n  of Neural Features", "abstract": "  We consider the problem of category-level 6D pose estimation from a single\nRGB image. Our approach represents an object category as a cuboid mesh and\nlearns a generative model of the neural feature activations at each mesh vertex\nto perform pose estimation through differentiable rendering. A common problem\nof rendering-based approaches is that they rely on bounding box proposals,\nwhich do not convey information about the 3D rotation of the object and are not\nreliable when objects are partially occluded. Instead, we introduce a\ncoarse-to-fine optimization strategy that utilizes the rendering process to\nestimate a sparse set of 6D object proposals, which are subsequently refined\nwith gradient-based optimization. The key to enabling the convergence of our\napproach is a neural feature representation that is trained to be scale- and\nrotation-invariant using contrastive learning. Our experiments demonstrate an\nenhanced category-level 6D pose estimation performance compared to prior work,\nparticularly under strong partial occlusion.\n", "date": "2022-09-12", "categories": []}
{"id": "2208.12986", "title": "6D Robotic Assembly Based on RGB-only Object Pose Estimation", "abstract": "  Vision-based robotic assembly is a crucial yet challenging task as the\ninteraction with multiple objects requires high levels of precision. In this\npaper, we propose an integrated 6D robotic system to perceive, grasp,\nmanipulate and assemble blocks with tight tolerances. Aiming to provide an\noff-the-shelf RGB-only solution, our system is built upon a monocular 6D object\npose estimation network trained solely with synthetic images leveraging\nphysically-based rendering. Subsequently, pose-guided 6D transformation along\nwith collision-free assembly is proposed to construct any designed structure\nwith arbitrary initial poses. Our novel 3-axis calibration operation further\nenhances the precision and robustness by disentangling 6D pose estimation and\nrobotic assembly. Both quantitative and qualitative results demonstrate the\neffectiveness of our proposed 6D robotic assembly system.\n", "date": "2022-08-27", "categories": []}
{"id": "2208.04052", "title": "Dataset of Industrial Metal Objects", "abstract": "  We present a diverse dataset of industrial metal objects. These objects are\nsymmetric, textureless and highly reflective, leading to challenging conditions\nnot captured in existing datasets. Our dataset contains both real-world and\nsynthetic multi-view RGB images with 6D object pose labels. Real-world data is\nobtained by recording multi-view images of scenes with varying object shapes,\nmaterials, carriers, compositions and lighting conditions. This results in over\n30,000 images, accurately labelled using a new public tool. Synthetic data is\nobtained by carefully simulating real-world conditions and varying them in a\ncontrolled and realistic way. This leads to over 500,000 synthetic images. The\nclose correspondence between synthetic and real-world data, and controlled\nvariations, will facilitate sim-to-real research. Our dataset's size and\nchallenging nature will facilitate research on various computer vision tasks\ninvolving reflective materials. The dataset and accompanying resources are made\navailable on the project website at https://pderoovere.github.io/dimo.\n", "date": "2022-08-08", "categories": []}
{"id": "2208.09829", "title": "CenDerNet: Center and Curvature Representations for Render-and-Compare\n  6D Pose Estimation", "abstract": "  We introduce CenDerNet, a framework for 6D pose estimation from multi-view\nimages based on center and curvature representations. Finding precise poses for\nreflective, textureless objects is a key challenge for industrial robotics. Our\napproach consists of three stages: First, a fully convolutional neural network\npredicts center and curvature heatmaps for each view; Second, center heatmaps\nare used to detect object instances and find their 3D centers; Third, 6D object\nposes are estimated using 3D centers and curvature heatmaps. By jointly\noptimizing poses across views using a render-and-compare approach, our method\nnaturally handles occlusions and object symmetries. We show that CenDerNet\noutperforms previous methods on two industry-relevant datasets: DIMO and\nT-LESS.\n", "date": "2022-08-21", "categories": []}
{"id": "2208.08807", "title": "COPE: End-to-end trainable Constant Runtime Object Pose Estimation", "abstract": "  State-of-the-art object pose estimation handles multiple instances in a test\nimage by using multi-model formulations: detection as a first stage and then\nseparately trained networks per object for 2D-3D geometric correspondence\nprediction as a second stage. Poses are subsequently estimated using the\nPerspective-n-Points algorithm at runtime. Unfortunately, multi-model\nformulations are slow and do not scale well with the number of object instances\ninvolved. Recent approaches show that direct 6D object pose estimation is\nfeasible when derived from the aforementioned geometric correspondences. We\npresent an approach that learns an intermediate geometric representation of\nmultiple objects to directly regress 6D poses of all instances in a test image.\nThe inherent end-to-end trainability overcomes the requirement of separately\nprocessing individual object instances. By calculating the mutual\nIntersection-over-Unions, pose hypotheses are clustered into distinct\ninstances, which achieves negligible runtime overhead with respect to the\nnumber of object instances. Results on multiple challenging standard datasets\nshow that the pose estimation performance is superior to single-model\nstate-of-the-art approaches despite being more than ~35 times faster. We\nadditionally provide an analysis showing real-time applicability (>24 fps) for\nimages where more than 90 object instances are present. Further results show\nthe advantage of supervising geometric-correspondence-based object pose\nestimation with the 6D pose.\n", "date": "2022-08-18", "categories": []}
{"id": "2208.10002", "title": "TransNet: Category-Level Transparent Object Pose Estimation", "abstract": "  Transparent objects present multiple distinct challenges to visual perception\nsystems. First, their lack of distinguishing visual features makes transparent\nobjects harder to detect and localize than opaque objects. Even humans find\ncertain transparent surfaces with little specular reflection or refraction,\ne.g. glass doors, difficult to perceive. A second challenge is that common\ndepth sensors typically used for opaque object perception cannot obtain\naccurate depth measurements on transparent objects due to their unique\nreflective properties. Stemming from these challenges, we observe that\ntransparent object instances within the same category (e.g. cups) look more\nsimilar to each other than to ordinary opaque objects of that same category.\nGiven this observation, the present paper sets out to explore the possibility\nof category-level transparent object pose estimation rather than instance-level\npose estimation. We propose TransNet, a two-stage pipeline that learns to\nestimate category-level transparent object pose using localized depth\ncompletion and surface normal estimation. TransNet is evaluated in terms of\npose estimation accuracy on a recent, large-scale transparent object dataset\nand compared to a state-of-the-art category-level pose estimation approach.\nResults from this comparison demonstrate that TransNet achieves improved pose\nestimation accuracy on transparent objects and key findings from the included\nablation studies suggest future directions for performance improvements.\n", "date": "2022-08-22", "categories": []}
{"id": "2208.08338", "title": "SO(3)-Pose: SO(3)-Equivariance Learning for 6D Object Pose Estimation", "abstract": "  6D pose estimation of rigid objects from RGB-D images is crucial for object\ngrasping and manipulation in robotics. Although RGB channels and the depth (D)\nchannel are often complementary, providing respectively the appearance and\ngeometry information, it is still non-trivial how to fully benefit from the two\ncross-modal data. From the simple yet new observation, when an object rotates,\nits semantic label is invariant to the pose while its keypoint offset direction\nis variant to the pose. To this end, we present SO(3)-Pose, a new\nrepresentation learning network to explore SO(3)-equivariant and\nSO(3)-invariant features from the depth channel for pose estimation. The\nSO(3)-invariant features facilitate to learn more distinctive representations\nfor segmenting objects with similar appearance from RGB channels. The\nSO(3)-equivariant features communicate with RGB features to deduce the (missed)\ngeometry for detecting keypoints of an object with the reflective surface from\nthe depth channel. Unlike most of existing pose estimation methods, our\nSO(3)-Pose not only implements the information communication between the RGB\nand depth channels, but also naturally absorbs the SO(3)-equivariance geometry\nknowledge from depth images, leading to better appearance and geometry\nrepresentation learning. Comprehensive experiments show that our method\nachieves the state-of-the-art performance on three benchmarks.\n", "date": "2022-08-17", "categories": []}
{"id": "2208.06661", "title": "SSP-Pose: Symmetry-Aware Shape Prior Deformation for Direct\n  Category-Level Object Pose Estimation", "abstract": "  Category-level pose estimation is a challenging problem due to intra-class\nshape variations. Recent methods deform pre-computed shape priors to map the\nobserved point cloud into the normalized object coordinate space and then\nretrieve the pose via post-processing, i.e., Umeyama's Algorithm. The\nshortcomings of this two-stage strategy lie in two aspects: 1) The surrogate\nsupervision on the intermediate results can not directly guide the learning of\npose, resulting in large pose error after post-processing. 2) The inference\nspeed is limited by the post-processing step. In this paper, to handle these\nshortcomings, we propose an end-to-end trainable network SSP-Pose for\ncategory-level pose estimation, which integrates shape priors into a direct\npose regression network. SSP-Pose stacks four individual branches on a shared\nfeature extractor, where two branches are designed to deform and match the\nprior model with the observed instance, and the other two branches are applied\nfor directly regressing the totally 9 degrees-of-freedom pose and performing\nsymmetry reconstruction and point-wise inlier mask prediction respectively.\nConsistency loss terms are then naturally exploited to align the outputs of\ndifferent branches and promote the performance. During inference, only the\ndirect pose regression branch is needed. In this manner, SSP-Pose not only\nlearns category-level pose-sensitive characteristics to boost performance but\nalso keeps a real-time inference speed. Moreover, we utilize the symmetry\ninformation of each category to guide the shape prior deformation, and propose\na novel symmetry-aware loss to mitigate the matching ambiguity. Extensive\nexperiments on public datasets demonstrate that SSP-Pose produces superior\nperformance compared with competitors with a real-time inference speed at about\n25Hz.\n", "date": "2022-08-13", "categories": []}
{"id": "2208.01172", "title": "MV6D: Multi-View 6D Pose Estimation on RGB-D Frames Using a Deep\n  Point-wise Voting Network", "abstract": "  Estimating 6D poses of objects is an essential computer vision task. However,\nmost conventional approaches rely on camera data from a single perspective and\ntherefore suffer from occlusions. We overcome this issue with our novel\nmulti-view 6D pose estimation method called MV6D which accurately predicts the\n6D poses of all objects in a cluttered scene based on RGB-D images from\nmultiple perspectives. We base our approach on the PVN3D network that uses a\nsingle RGB-D image to predict keypoints of the target objects. We extend this\napproach by using a combined point cloud from multiple views and fusing the\nimages from each view with a DenseFusion layer. In contrast to current\nmulti-view pose detection networks such as CosyPose, our MV6D can learn the\nfusion of multiple perspectives in an end-to-end manner and does not require\nmultiple prediction stages or subsequent fine tuning of the prediction.\nFurthermore, we present three novel photorealistic datasets of cluttered scenes\nwith heavy occlusions. All of them contain RGB-D images from multiple\nperspectives and the ground truth for instance semantic segmentation and 6D\npose estimation. MV6D significantly outperforms the state-of-the-art in\nmulti-view 6D pose estimation even in cases where the camera poses are known\ninaccurately. Furthermore, we show that our approach is robust towards dynamic\ncamera setups and that its accuracy increases incrementally with an increasing\nnumber of perspectives.\n", "date": "2022-08-01", "categories": []}
{"id": "2208.00113", "title": "Neural Correspondence Field for Object Pose Estimation", "abstract": "  We propose a method for estimating the 6DoF pose of a rigid object with an\navailable 3D model from a single RGB image. Unlike classical\ncorrespondence-based methods which predict 3D object coordinates at pixels of\nthe input image, the proposed method predicts 3D object coordinates at 3D query\npoints sampled in the camera frustum. The move from pixels to 3D points, which\nis inspired by recent PIFu-style methods for 3D reconstruction, enables\nreasoning about the whole object, including its (self-)occluded parts. For a 3D\nquery point associated with a pixel-aligned image feature, we train a\nfully-connected neural network to predict: (i) the corresponding 3D object\ncoordinates, and (ii) the signed distance to the object surface, with the first\ndefined only for query points in the surface vicinity. We call the mapping\nrealized by this network as Neural Correspondence Field. The object pose is\nthen robustly estimated from the predicted 3D-3D correspondences by the\nKabsch-RANSAC algorithm. The proposed method achieves state-of-the-art results\non three BOP datasets and is shown superior especially in challenging cases\nwith occlusion. The project website is at: linhuang17.github.io/NCF.\n", "date": "2022-07-30", "categories": []}
{"id": "2207.13264", "title": "Instance-specific 6-DoF Object Pose Estimation from Minimal Annotations", "abstract": "  In many robotic applications, the environment setting in which the 6-DoF pose\nestimation of a known, rigid object and its subsequent grasping is to be\nperformed, remains nearly unchanging and might even be known to the robot in\nadvance. In this paper, we refer to this problem as instance-specific pose\nestimation: the robot is expected to estimate the pose with a high degree of\naccuracy in only a limited set of familiar scenarios. Minor changes in the\nscene, including variations in lighting conditions and background appearance,\nare acceptable but drastic alterations are not anticipated. To this end, we\npresent a method to rapidly train and deploy a pipeline for estimating the\ncontinuous 6-DoF pose of an object from a single RGB image. The key idea is to\nleverage known camera poses and rigid body geometry to partially automate the\ngeneration of a large labeled dataset. The dataset, along with sufficient\ndomain randomization, is then used to supervise the training of deep neural\nnetworks for predicting semantic keypoints. Experimentally, we demonstrate the\nconvenience and effectiveness of our proposed method to accurately estimate\nobject pose requiring only a very small amount of manual annotation for\ntraining.\n", "date": "2022-07-27", "categories": []}
{"id": "2207.13691", "title": "ShAPO: Implicit Representations for Multi-Object Shape, Appearance, and\n  Pose Optimization", "abstract": "  Our method studies the complex task of object-centric 3D understanding from a\nsingle RGB-D observation. As it is an ill-posed problem, existing methods\nsuffer from low performance for both 3D shape and 6D pose and size estimation\nin complex multi-object scenarios with occlusions. We present ShAPO, a method\nfor joint multi-object detection, 3D textured reconstruction, 6D object pose\nand size estimation. Key to ShAPO is a single-shot pipeline to regress shape,\nappearance and pose latent codes along with the masks of each object instance,\nwhich is then further refined in a sparse-to-dense fashion. A novel\ndisentangled shape and appearance database of priors is first learned to embed\nobjects in their respective shape and appearance space. We also propose a\nnovel, octree-based differentiable optimization step, allowing us to further\nimprove object shape, pose and appearance simultaneously under the learned\nlatent space, in an analysis-by-synthesis fashion. Our novel joint implicit\ntextured object representation allows us to accurately identify and reconstruct\nnovel unseen objects without having access to their 3D meshes. Through\nextensive experiments, we show that our method, trained on simulated indoor\nscenes, accurately regresses the shape, appearance and pose of novel objects in\nthe real-world with minimal fine-tuning. Our method significantly out-performs\nall baselines on the NOCS dataset with an 8% absolute improvement in mAP for 6D\npose estimation. Project page:\nhttps://zubair-irshad.github.io/projects/ShAPO.html\n", "date": "2022-07-27", "categories": []}
{"id": "2207.05444", "title": "Category-Level 6D Object Pose and Size Estimation using Self-Supervised\n  Deep Prior Deformation Networks", "abstract": "  It is difficult to precisely annotate object instances and their semantics in\n3D space, and as such, synthetic data are extensively used for these tasks,\ne.g., category-level 6D object pose and size estimation. However, the easy\nannotations in synthetic domains bring the downside effect of synthetic-to-real\n(Sim2Real) domain gap. In this work, we aim to address this issue in the task\nsetting of Sim2Real, unsupervised domain adaptation for category-level 6D\nobject pose and size estimation. We propose a method that is built upon a novel\nDeep Prior Deformation Network, shortened as DPDN. DPDN learns to deform\nfeatures of categorical shape priors to match those of object observations, and\nis thus able to establish deep correspondence in the feature space for direct\nregression of object poses and sizes. To reduce the Sim2Real domain gap, we\nformulate a novel self-supervised objective upon DPDN via consistency learning;\nmore specifically, we apply two rigid transformations to each object\nobservation in parallel, and feed them into DPDN respectively to yield dual\nsets of predictions; on top of the parallel learning, an inter-consistency term\nis employed to keep cross consistency between dual predictions for improving\nthe sensitivity of DPDN to pose changes, while individual intra-consistency\nones are used to enforce self-adaptation within each learning itself. We train\nDPDN on both training sets of the synthetic CAMERA25 and real-world REAL275\ndatasets; our results outperform the existing methods on REAL275 test set under\nboth the unsupervised and supervised settings. Ablation studies also verify the\nefficacy of our designs. Our code is released publicly at\nhttps://github.com/JiehongLin/Self-DPDN.\n", "date": "2022-07-12", "categories": []}
{"id": "2207.07869", "title": "CA-SpaceNet: Counterfactual Analysis for 6D Pose Estimation in Space", "abstract": "  Reliable and stable 6D pose estimation of uncooperative space objects plays\nan essential role in on-orbit servicing and debris removal missions.\nConsidering that the pose estimator is sensitive to background interference,\nthis paper proposes a counterfactual analysis framework named CASpaceNet to\ncomplete robust 6D pose estimation of the spaceborne targets under complicated\nbackground. Specifically, conventional methods are adopted to extract the\nfeatures of the whole image in the factual case. In the counterfactual case, a\nnon-existent image without the target but only the background is imagined. Side\neffect caused by background interference is reduced by counterfactual analysis,\nwhich leads to unbiased prediction in final results. In addition, we also carry\nout lowbit-width quantization for CA-SpaceNet and deploy part of the framework\nto a Processing-In-Memory (PIM) accelerator on FPGA. Qualitative and\nquantitative results demonstrate the effectiveness and efficiency of our\nproposed method. To our best knowledge, this paper applies causal inference and\nnetwork quantization to the 6D pose estimation of space-borne targets for the\nfirst time. The code is available at\nhttps://github.com/Shunli-Wang/CA-SpaceNet.\n", "date": "2022-07-16", "categories": []}
{"id": "2207.08082", "title": "CATRE: Iterative Point Clouds Alignment for Category-level Object Pose\n  Refinement", "abstract": "  While category-level 9DoF object pose estimation has emerged recently,\nprevious correspondence-based or direct regression methods are both limited in\naccuracy due to the huge intra-category variances in object shape and color,\netc. Orthogonal to them, this work presents a category-level object pose and\nsize refiner CATRE, which is able to iteratively enhance pose estimate from\npoint clouds to produce accurate results. Given an initial pose estimate, CATRE\npredicts a relative transformation between the initial pose and ground truth by\nmeans of aligning the partially observed point cloud and an abstract shape\nprior. In specific, we propose a novel disentangled architecture being aware of\nthe inherent distinctions between rotation and translation/size estimation.\nExtensive experiments show that our approach remarkably outperforms\nstate-of-the-art methods on REAL275, CAMERA25, and LM benchmarks up to a speed\nof ~85.32Hz, and achieves competitive results on category-level tracking. We\nfurther demonstrate that CATRE can perform pose refinement on unseen category.\nCode and trained models are available.\n", "date": "2022-07-17", "categories": []}
{"id": "2207.04880", "title": "SDFEst: Categorical Pose and Shape Estimation of Objects from RGB-D\n  using Signed Distance Fields", "abstract": "  Rich geometric understanding of the world is an important component of many\nrobotic applications such as planning and manipulation. In this paper, we\npresent a modular pipeline for pose and shape estimation of objects from RGB-D\nimages given their category. The core of our method is a generative shape\nmodel, which we integrate with a novel initialization network and a\ndifferentiable renderer to enable 6D pose and shape estimation from a single or\nmultiple views. We investigate the use of discretized signed distance fields as\nan efficient shape representation for fast analysis-by-synthesis optimization.\nOur modular framework enables multi-view optimization and extensibility. We\ndemonstrate the benefits of our approach over state-of-the-art methods in\nseveral experiments on both synthetic and real data. We open-source our\napproach at https://github.com/roym899/sdfest.\n", "date": "2022-07-11", "categories": []}
{"id": "2207.02805", "title": "DPODv2: Dense Correspondence-Based 6 DoF Pose Estimation", "abstract": "  We propose a three-stage 6 DoF object detection method called DPODv2 (Dense\nPose Object Detector) that relies on dense correspondences. We combine a 2D\nobject detector with a dense correspondence estimation network and a multi-view\npose refinement method to estimate a full 6 DoF pose. Unlike other deep\nlearning methods that are typically restricted to monocular RGB images, we\npropose a unified deep learning network allowing different imaging modalities\nto be used (RGB or Depth). Moreover, we propose a novel pose refinement method,\nthat is based on differentiable rendering. The main concept is to compare\npredicted and rendered correspondences in multiple views to obtain a pose which\nis consistent with predicted correspondences in all views. Our proposed method\nis evaluated rigorously on different data modalities and types of training data\nin a controlled setup. The main conclusions is that RGB excels in\ncorrespondence estimation, while depth contributes to the pose accuracy if good\n3D-3D correspondences are available. Naturally, their combination achieves the\noverall best performance. We perform an extensive evaluation and an ablation\nstudy to analyze and validate the results on several challenging datasets.\nDPODv2 achieves excellent results on all of them while still remaining fast and\nscalable independent of the used data modality and the type of training data\n", "date": "2022-07-06", "categories": []}
{"id": "2207.02811", "title": "Multi-View Object Pose Refinement With Differentiable Renderer", "abstract": "  This paper introduces a novel multi-view 6 DoF object pose refinement\napproach focusing on improving methods trained on synthetic data. It is based\non the DPOD detector, which produces dense 2D-3D correspondences between the\nmodel vertices and the image pixels in each frame. We have opted for the use of\nmultiple frames with known relative camera transformations, as it allows\nintroduction of geometrical constraints via an interpretable ICP-like loss\nfunction. The loss function is implemented with a differentiable renderer and\nis optimized iteratively. We also demonstrate that a full detection and\nrefinement pipeline, which is trained solely on synthetic data, can be used for\nauto-labeling real data. We perform quantitative evaluation on LineMOD,\nOcclusion, Homebrewed and YCB-V datasets and report excellent performance in\ncomparison to the state-of-the-art methods trained on the synthetic and real\ndata. We demonstrate empirically that our approach requires only a few frames\nand is robust to close camera locations and noise in extrinsic camera\ncalibration, making its practical usage easier and more ubiquitous.\n", "date": "2022-07-06", "categories": []}
{"id": "2207.00260", "title": "Towards Two-view 6D Object Pose Estimation: A Comparative Study on\n  Fusion Strategy", "abstract": "  Current RGB-based 6D object pose estimation methods have achieved noticeable\nperformance on datasets and real world applications. However, predicting 6D\npose from single 2D image features is susceptible to disturbance from changing\nof environment and textureless or resemblant object surfaces. Hence, RGB-based\nmethods generally achieve less competitive results than RGBD-based methods,\nwhich deploy both image features and 3D structure features. To narrow down this\nperformance gap, this paper proposes a framework for 6D object pose estimation\nthat learns implicit 3D information from 2 RGB images. Combining the learned 3D\ninformation and 2D image features, we establish more stable correspondence\nbetween the scene and the object models. To seek for the methods best utilizing\n3D information from RGB inputs, we conduct an investigation on three different\napproaches, including Early- Fusion, Mid-Fusion, and Late-Fusion. We ascertain\nthe Mid- Fusion approach is the best approach to restore the most precise 3D\nkeypoints useful for object pose estimation. The experiments show that our\nmethod outperforms state-of-the-art RGB-based methods, and achieves comparable\nresults with RGBD-based methods.\n", "date": "2022-07-01", "categories": []}
{"id": "2206.15436", "title": "Category-Level 6D Object Pose Estimation in the Wild: A Semi-Supervised\n  Learning Approach and A New Dataset", "abstract": "  6D object pose estimation is one of the fundamental problems in computer\nvision and robotics research. While a lot of recent efforts have been made on\ngeneralizing pose estimation to novel object instances within the same\ncategory, namely category-level 6D pose estimation, it is still restricted in\nconstrained environments given the limited number of annotated data. In this\npaper, we collect Wild6D, a new unlabeled RGBD object video dataset with\ndiverse instances and backgrounds. We utilize this data to generalize\ncategory-level 6D object pose estimation in the wild with semi-supervised\nlearning. We propose a new model, called Rendering for Pose estimation network\nRePoNet, that is jointly trained using the free ground-truths with the\nsynthetic data, and a silhouette matching objective function on the real-world\ndata. Without using any 3D annotations on real data, our method outperforms\nstate-of-the-art methods on the previous dataset and our Wild6D test set (with\nmanual annotations for evaluation) by a large margin. Project page with Wild6D\ndata: https://oasisyang.github.io/semi-pose .\n", "date": "2022-06-30", "categories": []}
{"id": "2206.11808", "title": "Unseen Object 6D Pose Estimation: A Benchmark and Baselines", "abstract": "  Estimating the 6D pose for unseen objects is in great demand for many\nreal-world applications. However, current state-of-the-art pose estimation\nmethods can only handle objects that are previously trained. In this paper, we\npropose a new task that enables and facilitates algorithms to estimate the 6D\npose estimation of novel objects during testing. We collect a dataset with both\nreal and synthetic images and up to 48 unseen objects in the test set. In the\nmean while, we propose a new metric named Infimum ADD (IADD) which is an\ninvariant measurement for objects with different types of pose ambiguity. A\ntwo-stage baseline solution for this task is also provided. By training an\nend-to-end 3D correspondences network, our method finds corresponding points\nbetween an unseen object and a partial view RGBD image accurately and\nefficiently. It then calculates the 6D pose from the correspondences using an\nalgorithm robust to object symmetry. Extensive experiments show that our method\noutperforms several intuitive baselines and thus verify its effectiveness. All\nthe data, code and models will be made publicly available. Project page:\nwww.graspnet.net/unseen6d\n", "date": "2022-06-23", "categories": []}
{"id": "2206.01831", "title": "Spatial Feature Mapping for 6DoF Object Pose Estimation", "abstract": "  This work aims to estimate 6Dof (6D) object pose in background clutter.\nConsidering the strong occlusion and background noise, we propose to utilize\nthe spatial structure for better tackling this challenging task. Observing that\nthe 3D mesh can be naturally abstracted by a graph, we build the graph using 3D\npoints as vertices and mesh connections as edges. We construct the\ncorresponding mapping from 2D image features to 3D points for filling the graph\nand fusion of the 2D and 3D features. Afterward, a Graph Convolutional Network\n(GCN) is applied to help the feature exchange among objects' points in 3D\nspace. To address the problem of rotation symmetry ambiguity for objects, a\nspherical convolution is utilized and the spherical features are combined with\nthe convolutional features that are mapped to the graph. Predefined 3D\nkeypoints are voted and the 6DoF pose is obtained via the fitting optimization.\nTwo scenarios of inference, one with the depth information and the other\nwithout it are discussed. Tested on the datasets of YCB-Video and LINEMOD, the\nexperiments demonstrate the effectiveness of our proposed method.\n", "date": "2022-06-03", "categories": []}
{"id": "2206.02531", "title": "3D-Augmented Contrastive Knowledge Distillation for Image-based Object\n  Pose Estimation", "abstract": "  Image-based object pose estimation sounds amazing because in real\napplications the shape of object is oftentimes not available or not easy to\ntake like photos. Although it is an advantage to some extent, un-explored shape\ninformation in 3D vision learning problem looks like \"flaws in jade\". In this\npaper, we deal with the problem in a reasonable new setting, namely 3D shape is\nexploited in the training process, and the testing is still purely image-based.\nWe enhance the performance of image-based methods for category-agnostic object\npose estimation by exploiting 3D knowledge learned by a multi-modal method.\nSpecifically, we propose a novel contrastive knowledge distillation framework\nthat effectively transfers 3D-augmented image representation from a multi-modal\nmodel to an image-based model. We integrate contrastive learning into the\ntwo-stage training procedure of knowledge distillation, which formulates an\nadvanced solution to combine these two approaches for cross-modal tasks. We\nexperimentally report state-of-the-art results compared with existing\ncategory-agnostic image-based methods by a large margin (up to +5% improvement\non ObjectNet3D dataset), demonstrating the effectiveness of our method.\n", "date": "2022-06-02", "categories": []}
{"id": "2606.22076", "title": "Learning Cross-View Semantic Priors for Single-Reference Unseen Object Pose Estimation", "abstract": "Single-reference unseen object 6D pose estimation reduces object onboarding by estimating poses of arbitrary novel objects from only one reference view. Recent correspondence-based pipelines have achieved robust performance with vision foundation model (VFM) features. However, they typically treat these features as intra-view descriptors, leaving dense visual-semantic cues, including appearance, structure, and context, insufficiently exchanged across views before geometric decoding. Consequently, the decoded point features may lack joint semantic and geometric discriminability, making correspondence estimation still difficult in challenging cases. Instead of processing features independently, we build the correspondence pipeline around an early cross-view semantic prior. Specifically, cross-view semantic interaction (CVSI) enables dense query and reference VFM tokens to exchange semantic context and form a cross-view prior. Nevertheless, direct CVSI may disturb the VFM token structure, while the resulting semantic prior still needs 3D representation consistency for rigid correspondence. To make this CVSI prior reliable for 3D correspondence learning, we introduce two complementary training-time constraints: the intra-view structure preservation (IVSP) loss preserves the original intra-view token affinity structure during interaction, while the reference-anchored geometric consistency (RAGC) loss enforces spatial representation consistency of decoded point features. The final pose is recovered from learned correspondences through weighted SVD. We further construct a challenging view-pair protocol from the BOP Challenge datasets YCB-V and TUD-L to evaluate robustness in difficult matching scenarios. Extensive experiments on six benchmarks under different view-pair settings show that our method achieves state-of-the-art performance while maintaining comparable inference speed.", "date": "2026-06-20", "categories": []}
{"id": "2606.23293", "title": "Flow6D: Discrete-to-Continuous Flow Matching for Efficient and Accurate Category-Level 6D Pose Estimation", "abstract": "6D pose estimation is a key task in computer vision and embodied AI, widely used in robotic manipulation, augmented reality, etc. Existing methods directly regress in a high-dimensional continuous space, facing two key challenges in category-level pose estimation: limited accuracy due to noise and local optima, and inefficient search over an infinite space that hinders real-time performance. This paper proposes Flow6D, a hierarchical flow matching framework with a two-stage discrete latent space localization-continuous pose regression strategy. Rotation and translation parameters are first discretized into bins, with a discrete flow matching model locking the latent space around the true pose to reduce search complexity. Then, by sampling in the latent space, a continuous flow matching model predicts local pose residuals to optimize the estimate and regress to an accurate pose. The framework also naturally extends to articulated objects, outperforming state-of-the-art methods on synthetic and real datasets with real-time inference at 70 FPS. Project website: https://flow6d.github.io/.", "date": "2026-06-22", "categories": []}
{"id": "2606.23634", "title": "Pose Anything Anywhere:Model-free Object Poses from Arbitrary References", "abstract": "Estimating the 6D pose of unseen objects is a fundamental yet challenging problem for open-world robotics and embodied perception. Model-based methods are accurate but depend on CAD assets or heavy onboarding, while most model-free approaches are still limited to pairwise single-anchor matching and thus fail under occlusion and large viewpoint changes with low query-reference overlap. Therefore, we present PANY, a unified model-free framework that seamlessly supports both RGB and RGB-D inputs, operates on one or sparse pose-free reference views, and generalizes effectively to novel objects. Built on a multi-view transformer geometry backbone, PANY moves beyond pairwise matching by learning view-consistent geometry and cross-view alignment cues that remain stable under wide baselines and limited overlap. When additional unposed assist views are available, PANY aggregates them via pose-graph canonical registration to increase geometric coverage and reinforce the final pose. Extensive experiments show that PANY achieves state-of-the-art performance across multiple benchmarks, substantially outperforming existing model-free methods, improving pose accuracy by +12% on YCB-V and over +20% on LM-O. Furthermore, PANY consistently performs well under both single-reference and sparse-reference settings, demonstrating strong robustness in real-world environments.", "date": "2026-06-22", "categories": []}
{"id": "2606.29697", "title": "MF-UAVPose6D: A Model-Free Monocular 6-DoF Pose Estimation Framework for Fixed-Wing UAVs", "abstract": "For uncrewed aerial vehicles (UAVs), estimating six-degree-of-freedom (6-DoF) poses is essential for airspace situational awareness, target tracking, and counter-UAV operations. However, non-cooperative targets usually lack computer-aided design (CAD) models and keypoint priors, making existing model-based or keypoint-matching methods difficult to apply reliably. To address these challenges, this paper proposes MF-UAVPose6D, a model-free monocular 6-DoF pose estimation framework for fixed-wing UAVs. During inference, the method takes only a single red-green-blue (RGB) image and camera intrinsics as input. It first obtains a stable target anchor through heatmap-guided center localization, introduces a Perspective-Aware Module (PAM) to model observation-ray priors, exploits Dynamic Topological Sampling (DTS) to complement weak structural cues from the wings, fuselage, and tail, and adopts a decoupled translation-rotation pose decoding mechanism to estimate the 6-DoF pose. In addition, we construct the FW-UAV6DPose synthetic dataset, which covers fixed-wing UAV observations across diverse distances, viewpoints, and poses. Experimental results show that MF-UAVPose6D achieves accurate and efficient monocular 6-DoF pose estimation without requiring CAD models, and demonstrates strong robustness in long-range rotation estimation, depth recovery, and joint pose evaluation.", "date": "2026-06-29", "categories": []}
{"id": "2606.30058", "title": "Emergence of a Shared Canonical Object Frame from In-the-Wild Videos", "abstract": "Comparing object orientations and positions across different instances requires their poses to be expressed in a shared canonical frame. Establishing such frames has traditionally required manual annotation, creating a scaling bottleneck that limits category and instance diversity. We show that a shared canonical frame can instead emerge from self-supervised training on object-centric videos captured in the wild, using only noisy camera poses from Structure-from-Motion. Our key idea is to route all training sequences through a shared geometric bottleneck: a coarse canonical mesh that carries no category-specific detail. By learning dense correspondences from image pixels to this mesh, and estimating per-sequence alignments from noisy SfM geometry, a common canonical frame emerges from multi-view consistency and the semantic priors of the feature extractor, without any canonical pose labels or category conditioning. Trained in a self-supervised manner on 160,000 in-the-wild object videos, our method achieves competitive accuracy on category-level pose estimation benchmarks compared to methods that rely on canonical pose supervision. The code and checkpoint is available on https://github.com/Fischer-Tom/Emergent-Canonical-Frame/.", "date": "2026-06-29", "categories": []}