Image-to-3D

Add model card with pipeline tag, project links and checkpoints

#1
by nielsr HF Staff - opened
Files changed (1) hide show
  1. README.md +48 -3
README.md CHANGED
@@ -1,3 +1,48 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ pipeline_tag: image-to-3d
4
+ ---
5
+
6
+ # ViewSplat: View-Adaptive 3D Gaussian Splatting for Feed-Forward Synthesis
7
+
8
+ This repository contains the pre-trained checkpoints for **ViewSplat**, a view-adaptive 3D Gaussian splatting network for novel view synthesis from unposed images, presented at ECCV 2026.
9
+
10
+ [Paper](https://huggingface.co/papers/2603.25265) | [Project Page](https://cvlab-uos.github.io/ViewSplat) | [GitHub](https://github.com/cvlab-uos/ViewSplat)
11
+
12
+ <p align="center">
13
+ <img src="https://cvlab-uos.github.io/ViewSplat/static/images/intro_figure_v2.png" alt="Overview of ViewSplat" width="90%">
14
+ </p>
15
+
16
+ ## Introduction
17
+
18
+ ViewSplat shifts the paradigm of feed-forward 3D Gaussian splatting from static primitive regression to view-adaptive splatting. Instead of a rigid Gaussian representation, the pipeline learns a view-adaptive latent representation. ViewSplat initially predicts base Gaussian primitives alongside the weights of scene-conditioned View MLPs. During rendering, these MLPs take target-view coordinates as input and predict view-dependent residual updates for each Gaussian attribute (i.e., 3D position, scale, rotation, opacity, and color). This mechanism allows each primitive to rectify initial estimation errors, capturing high-fidelity appearances in real-time.
19
+
20
+ ## Pre-trained Checkpoints
21
+
22
+ The following pre-trained checkpoints are hosted in this repository:
23
+
24
+ | Model name | Training data | Training settings |
25
+ |:---:|:---:|:---:|
26
+ | [re10k_spf_viewsplat.ckpt](https://huggingface.co/myeon01/ViewSplat/resolve/main/re10k_spf_viewsplat.ckpt) | RE10K | 2 views, SPFSplat-based |
27
+ | [acid_spf_viewsplat.ckpt](https://huggingface.co/myeon01/ViewSplat/resolve/main/acid_spf_viewsplat.ckpt) | ACID | 2 views, SPFSplat-based |
28
+ | [re10k_spfv2_viewsplat.ckpt](https://huggingface.co/myeon01/ViewSplat/resolve/main/re10k_spfv2_viewsplat.ckpt) | RE10K | 2 views, SPFSplatV2-based |
29
+ | [acid_spfv2_viewsplat.ckpt](https://huggingface.co/myeon01/ViewSplat/resolve/main/acid_spfv2_viewsplat.ckpt) | ACID | 2 views, SPFSplatV2-based |
30
+ | [re10k_spfv2l_viewsplat.ckpt](https://huggingface.co/myeon01/ViewSplat/resolve/main/re10k_spfv2l_viewsplat.ckpt) | RE10K | 2 views, SPFSplatV2-L-based |
31
+ | [acid_spfv2l_viewsplat.ckpt](https://huggingface.co/myeon01/ViewSplat/resolve/main/acid_spfv2l_viewsplat.ckpt) | ACID | 2 views, SPFSplatV2-L-based |
32
+
33
+ ## Running the Code
34
+
35
+ For instructions on installation, dataset preparation, training, and evaluation, please refer to the official [GitHub Repository](https://github.com/cvlab-uos/ViewSplat).
36
+
37
+ ## Citation
38
+
39
+ If you find ViewSplat useful for your research, please cite:
40
+
41
+ ```bibtex
42
+ @article{Jeong2026viewsplat,
43
+ title={ViewSplat: View-Adaptive 3D Gaussian Splatting for Feed-Forward Synthesis},
44
+ author={Jeong, Moonyeon and Min, Seunggi tensors and Lee, Suhyeon and Seong, Hongje},
45
+ journal={arXiv preprint arXiv: 2603.25265},
46
+ year={2026}
47
+ }
48
+ ```