Release unified Body Pose, Hand Pose and native77 Motion (2026-10-20)

.gitattributes

@@ -33,9 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
-onnx/gem_denoiser.onnx.data filter=lfs diff=lfs merge=lfs -text
-onnx/gem_denoiser_no_imgfeat.onnx.data filter=lfs diff=lfs merge=lfs -text
 onnx/vitpose.onnx.data filter=lfs diff=lfs merge=lfs -text
-onnx/sam3db_backbone.onnx.data filter=lfs diff=lfs merge=lfs -text
-gem_smpl/onnx/vitpose_coco17.onnx.data filter=lfs diff=lfs merge=lfs -text
-gem_smpl/onnx/hmr2.onnx.data filter=lfs diff=lfs merge=lfs -text

README.md

@@ -22,6 +22,10 @@ All training data and the SOMA parametric body model are owned by NVIDIA or rele
 GEM was developed by NVIDIA.  
 *This model is ready for commercial or non-commercial use.*
 
+This card describes temporal Body Motion. The repository also distributes
+separate Body Pose and Hand Pose checkpoints for image inputs; the checkpoint
+bundle below identifies their configurations and roles.
+
 ### License/Terms of Use:
 
 Use of the source code is governed by the [Apache License, Version 2.0](https://www.apache.org/licenses/LICENSE-2.0). Use of the associated model is governed by the [NVIDIA Open Model License Agreement](https://www.nvidia.com/en-us/agreements/enterprise-software/nvidia-open-model-license/).
@@ -53,9 +57,9 @@ GEM is intended for use by individuals and professionals in fields such as machi
 ## Model Architecture:
 
 **Architecture Type:** Transformer  
-**Network Architecture:** 12-layer Transformer Decoder  
+**Network Architecture:** 12-layer temporal Transformer, hidden size 512, 8 attention heads<br>
 **This model was developed based on Not Applicable (N/A).**  
-**Number of model parameters:** 520M (5.2*10^8)  
+**Number of model parameters:** 46.8M total temporal Motion parameters (46,762,136), including 44.2M trainable parameters (44,202,136). This excludes the separate Body Pose feature extractor, Hand Pose model, and detectors.
 
 ## Computational Load
 
@@ -68,7 +72,7 @@ GEM is intended for use by individuals and professionals in fields such as machi
 
 **Input Format(s):**
 
-- Video: RGB tensor `(T, H, W, 3)`, any resolution, internally processed via the frozen video encoder
+- Video: RGB tensor `(T, H, W, 3)`, any resolution, internally processed via the frozen Body Pose image encoder
 - Other — person bounding box: `(T, 4)` float array in `xyxy` pixel coordinates
 - Other — camera intrinsics: `(T, 3, 3)` float array of per-frame camera intrinsic matrices containing focal lengths and principal-point coordinates
 
@@ -108,6 +112,42 @@ GEM v1.1
 An installable Python package provides the `gem-infer`, `gem-render`, `gem-train`, and `gem-eval` entry points. Body Motion inference uses `gem-infer motion-video --task body` for RGB video or `gem-infer motion-cache --task body` for prepared conditions. The package requires Python 3.10 or newer. The validated CUDA runtime uses PyTorch 2.10 and SOMA 0.3.3 with separately supplied model weights and geometry assets. Mesh/video rendering uses the documented x86/Open3D 0.19 environment.  
  
 
+## Pretrained Checkpoint Bundle
+
+The unified release provides three model checkpoints with matching public
+configurations. Body and Hand exports contain complete model weights for
+inference and fine-tuning initialization; optimizer and training-loop state are
+not included, so these files are not full training-resume checkpoints.
+
+| Component | Weights | Configuration / companion files |
+| --- | --- | --- |
+| Body Pose | [body_s240000.ckpt](https://huggingface.co/nvidia/GEM-X/resolve/main/body_s240000.ckpt) | [body_config.json](https://huggingface.co/nvidia/GEM-X/blob/main/body_config.json); preset `body` |
+| Hand Pose | [egohand_s060000.ckpt](https://huggingface.co/nvidia/GEM-X/resolve/main/egohand_s060000.ckpt) | [hand_config.json](https://huggingface.co/nvidia/GEM-X/blob/main/hand_config.json); preset `hand`; includes the learned RLE density used for fine-tuning |
+| Body Motion native77 | [motion_native77_e774_s292175.ckpt](https://huggingface.co/nvidia/GEM-X/resolve/main/motion_native77_e774_s292175.ckpt) | [motion_stats_raw.npz](https://huggingface.co/nvidia/GEM-X/resolve/main/motion_stats_raw.npz) and [motion_native77.yaml](https://huggingface.co/nvidia/GEM-X/blob/main/motion_native77.yaml) |
+| Original native77 ViTPose | [vitpose.pth](https://huggingface.co/nvidia/GEM-X/resolve/main/vitpose.pth) | Used by [body_video_native77.yaml](https://huggingface.co/nvidia/GEM-X/blob/main/body_video_native77.yaml) |
+
+The video recipe pairs the original final775 Motion weights with Body Pose
+s240000 features and the existing original ViTPose detector. The experimental
+random-blur epoch10 detector is not included in this release bundle.
+
+Download the model files together, preserving their relative placement:
+
+```bash
+hf download nvidia/GEM-X \
+  body_s240000.ckpt body_config.json \
+  egohand_s060000.ckpt hand_config.json \
+  motion_native77_e774_s292175.ckpt motion_stats_raw.npz \
+  motion_native77.yaml body_video_native77.yaml vitpose.pth \
+  --local-dir weights
+```
+
+The video YAML resolves checkpoint paths relative to its own directory. Obtain
+the YOLOX HumanArt person detector separately as `weights/yolox_humanart.onnx`,
+or change that path in the YAML. SOMA 0.3.3 geometry assets and an optional
+RT-DETR Hand detector are also supplied separately. Follow the
+[GEM-X installation and inference guides](https://github.com/NVlabs/GEM-X)
+for these prerequisites and the unified command-line interface.
+
 ## Training, Testing, and Evaluation Datasets:
 
 
@@ -147,7 +187,7 @@ An installable Python package provides the `gem-infer`, `gem-render`, `gem-train
 
 ## Inference:
 
-**Acceleration Engine:** TensorRT, PyTorch
+**Acceleration Engine:** PyTorch with CUDA
 **Hardware Requirements (GPU Architecture, Model):**  NVIDIA Ampere A100 80GB
 
 ## Ethical Considerations:

body_config.json

@@ -0,0 +1,42 @@
+{
+  "task": "body",
+  "head": {
+    "num_body_joints": 76,
+    "num_identity_coeffs": 128,
+    "num_scale_params": 57,
+    "scale_comps_dim": 0,
+    "scale_param_mode": "direct",
+    "scale_parameterization": "packed",
+    "fixed_global_scale": 1.0,
+    "transformer_dim": 1024,
+    "transformer_depth": 6,
+    "transformer_heads": 8,
+    "transformer_dim_head": 64,
+    "transformer_mlp_dim": 1024,
+    "transformer_context_dim": 1280,
+    "transformer_dropout": 0.0,
+    "transformer_emb_dropout": 0.0,
+    "ief_iters": 1,
+    "learnable_init": true,
+    "use_init_to_token": true,
+    "two_layer_heads": true,
+    "per_layer_predictions": true,
+    "cliff_cond_dim": 3,
+    "camera_scale_mode": "ste_clamp"
+  },
+  "backbone": {
+    "name": "dinov3_vith16plus",
+    "drop_path_rate": 0.1
+  },
+  "conditioning": {
+    "camera_mode": "sam3db_perspective"
+  },
+  "backbone_input_crop_w": null,
+  "ray_pe_num_bands": null,
+  "kp2d_translation": null,
+  "pose_release_id": "gem-pose-body-s240000-v1",
+  "inference": {
+    "camera_representation": "legacy",
+    "minimum_depth": 0.05
+  }
+}

gem_smpl.ckpt → body_s240000.ckpt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1d15cbe2864d6de61a75e83fdbfe83bec3c7b183eee3d3dcdbd9107e4456454a
-size 5518923941
+oid sha256:e244d27f2834eb6382c5f6b9cbf2a38681b685514a1b2e86f8f3d2592d614959
+size 3538753341

body_video_native77.yaml

@@ -0,0 +1,48 @@
+# Original native77 final775 Motion with original ViTPose and Body Pose features.
+# Checkpoint and statistics paths resolve relative to this file.
+# Obtain yolox_humanart.onnx separately and place it here or change its path.
+# Keep motion_stats_raw.npz; the codec uses its trained 0.01 std floor.
+pose:
+  checkpoint: body_s240000.ckpt
+  release: body
+  batch_size: 4
+observations:
+  checkpoint: vitpose.pth
+  release_id: vitpose77-dinov3-huge-metrosim-256x192-v1
+  batch_size: 4
+detector:
+  checkpoint: yolox_humanart.onnx
+  device: cpu
+motion:
+  checkpoint: motion_native77_e774_s292175.ckpt
+  stats: motion_stats_raw.npz
+  network:
+    output_dim: 655
+    max_len: 120
+    cliffcam_dim: 3
+    cam_angvel_dim: 6
+    imgseq_dim: 1024
+    latent_dim: 512
+    num_layers: 12
+    num_heads: 8
+    mlp_ratio: 4.0
+    pred_cam_dim: 3
+    static_conf_dim: 6
+    dropout: 0.1
+    avgbeta: true
+    obs_num_joints: 77
+    sequence_static_slices:
+    - - 456
+      - 584
+    - - 584
+      - 640
+  condition_spec:
+    pose_release_id: gem-pose-body-s240000-v1:pose_tokens
+    observation_source: vitpose77-dinov3-huge-metrosim-256x192-v1
+    observation_joint_set: soma77_native_v1
+  time_options:
+    reference_fps: 30.0
+    default_fps: 30.0
+    fps_by_data_name: {}
+camera:
+  mode: background

config.json

@@ -1,40 +0,0 @@
-{
-  "model_type": "gem",
-  "body_model": "soma",
-  "denoiser": {
-    "latent_dim": 1024,
-    "num_layers": 16,
-    "num_heads": 8,
-    "mlp_ratio": 4,
-    "scale_params_dim": 69,
-    "scale_comps_dim": 28,
-    "fix_scale_comps": true,
-    "encode_text": false,
-    "encoded_text_dim": 4096,
-    "text_encoder_cfg": {
-      "mode": "none",
-      "cross_attn_type": "mha"
-    },
-    "text_mask_prob": 0.1,
-    "music_mask_prob": 0.1,
-    "speech_mask_prob": 0.1,
-    "use_text_pos_enc": true
-  },
-  "endecoder": {
-    "stats_name": "MM_V2_SOMA_METROSIM",
-    "encode_type": "soma_v2",
-    "feat_dim": 585,
-    "clip_std": true
-  },
-  "diffusion": {
-    "sampler": "ddim",
-    "noise_schedule": "cosine",
-    "test_timestep_respacing": "50",
-    "ddim_eta": 0.0,
-    "guidance_param": 2.5
-  },
-  "pipeline": {
-    "remove_fingers": true,
-    "extract_features": false
-  }
-}

gem_smpl/missing_hmr4d_support/inputs/EMDB/hmr4d_support/emdb_slam_traj.pt → egohand_s060000.ckpt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0f760c4327b2cec66dae849b6345f2730f66f338285609c831d45d74560aa2b6
-size 5042492
+oid sha256:af56925fa08b7736787346c95b6bfe4fe7b857cb81e5803c336f72fb114eb900
+size 3552688469

gem_smpl/missing_hmr4d_support/README_GEM_missing_files.txt

@@ -1,19 +0,0 @@
-GEM missing hmr4d_support files
-
-This folder contains supplemental EMDB / RICH / 3DPW support files for GEM.
-
-Copy or merge the `inputs/` directory in this folder into the root of the
-GEM repository. After copying, the files should be located at:
-
-inputs/RICH/hmr4d_support/rich_test_vimo_preproc.pt
-inputs/EMDB/hmr4d_support/emdb_vimo.pt
-inputs/EMDB/hmr4d_support/emdb_slam_traj.pt
-inputs/3DPW/hmr4d_support/test_3dpw_vimo_labels.pt
-inputs/3DPW/hmr4d_support/3dpw_test_slam_traj.pt
-
-These paths are the ones used by:
-
-gem/datasets/rich/rich_motion_test.py
-gem/datasets/emdb/emdb_motion_test.py
-gem/datasets/threedpw/threedpw_motion_test.py
-gem/datasets/threedpw/threedpw_occ_motion_test.py

gem_smpl/missing_hmr4d_support/inputs/3DPW/hmr4d_support/3dpw_test_slam_traj.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:16340d0f2cf1453762163082f327d0a3cebd1488399dff71e60e1d001e53d249
-size 3271056

gem_smpl/missing_hmr4d_support/inputs/EMDB/hmr4d_support/emdb_vimo.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:c40713cf7b57e76a645edaa43c32d46aa6814daedaafbadaf1db16f5881c06e7
-size 999364858

gem_smpl/missing_hmr4d_support/inputs/RICH/hmr4d_support/rich_test_vimo_preproc.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4506ad4b0663b7933e4679f367b23067a06671f35c0c59a0bf1a9418a894296d
-size 1653984970

gem_smpl/onnx/gem_smpl_denoiser.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f174f92a07c40626e953bfec8c3c9c44ae94e2f68203348969b06654e69cf4f3
-size 1734170359

gem_smpl/onnx/gem_smpl_denoiser_no_imgfeat.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ad2255e3ec64f799a0191b4d92c45d5af21014ea71e3cff7923e3cb8312df914
-size 1734171175

gem_smpl/onnx/hmr2.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4f78023fd2abbaf26805b440c7e3c7a28187c920af389ddffc12e91c57a99f19
-size 724939

gem_smpl/onnx/hmr2.onnx.data

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:9ae90b247a5b5bbd1f4fba781a67eab03aa62ae6331d2ca55d7363b1106612b1
-size 2681192448

gem_smpl/onnx/vitpose_coco17.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:0c8d1cbf52ab358e5152b595b5ec0fe848d75c7b156291bc7af85d5f18c2bbcc
-size 341207

gem_smpl/onnx/vitpose_coco17.onnx.data

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:6521118c8c1d9deb29285ffd5a06bf6ab140a33d93c82daab6de843f236c6286
-size 2548839424

gem_smpl_config.json

@@ -1,33 +0,0 @@
-{
-  "model_type": "gem",
-  "body_model": "smplx",
-  "denoiser": {
-    "latent_dim": 1024,
-    "num_layers": 16,
-    "num_heads": 8,
-    "mlp_ratio": 4,
-    "encode_text": false,
-    "encoded_text_dim": 1024,
-    "text_mask_prob": 0.1,
-    "music_mask_prob": 0.1,
-    "speech_mask_prob": 0.1,
-    "use_text_pos_enc": true
-  },
-  "endecoder": {
-    "stats_name": "MM_V1_AMASS_LOCAL_BEDLAM_CAM",
-    "encode_type": "gvhmr",
-    "feat_dim": 151,
-    "clip_std": true
-  },
-  "diffusion": {
-    "sampler": "ddim",
-    "noise_schedule": "cosine",
-    "test_timestep_respacing": "50",
-    "ddim_eta": 0.0,
-    "guidance_param": 2.5
-  },
-  "pipeline": {
-    "remove_fingers": false,
-    "extract_features": false
-  }
-}

gem_soma.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4c1f85ca8c1e11e6588aead49fbc024bf660708def670043e0b537c101ee298e
-size 541758499

hand_config.json

@@ -0,0 +1,64 @@
+{
+  "task": "hand",
+  "head": {
+    "num_finger_joints": 24,
+    "num_identity_coeffs": 5,
+    "num_scale_params": 27,
+    "transformer_dim": 1024,
+    "transformer_depth": 6,
+    "transformer_heads": 8,
+    "transformer_dim_head": 64,
+    "transformer_mlp_dim": 1024,
+    "transformer_context_dim": 1280,
+    "transformer_dropout": 0.0,
+    "transformer_emb_dropout": 0.0,
+    "ief_iters": 1,
+    "learnable_init": true,
+    "use_init_to_token": true,
+    "two_layer_heads": true,
+    "per_layer_predictions": true,
+    "cliff_cond_dim": 3,
+    "ray_cond_dim": 32,
+    "camera_scale_mode": "ste_clamp",
+    "camera_depth_mode": "log_lambda",
+    "init_global_scale": 1.0,
+    "init_camera_scalar": -1.8199350824986398,
+    "camera_scalar_std": 0.3735116511306386
+  },
+  "backbone": {
+    "name": "dinov3_vith16plus",
+    "drop_path_rate": 0.0
+  },
+  "conditioning": {
+    "camera_mode": "sam3db_perspective"
+  },
+  "backbone_input_crop_w": null,
+  "ray_pe_num_bands": 8,
+  "kp2d_translation": {
+    "enabled": true,
+    "topology": "separate",
+    "num_joints": 25,
+    "d_model": 256,
+    "feat_hw": [
+      16,
+      16
+    ],
+    "decoder_depth": 3,
+    "num_heads": 8,
+    "presence_threshold": 0.5,
+    "negative_presence_weight": 1.0,
+    "min_points": 4,
+    "min_depth": 0.0001,
+    "damping": 1e-06,
+    "max_condition_number": 1000000.0
+  },
+  "pose_release_id": "gem-pose-hand-ga-s060000",
+  "inference": {
+    "hand_mirror": "full_image",
+    "camera_representation": "log_lambda",
+    "camera_log_value_bounds": [
+      -3.5,
+      3.0
+    ]
+  }
+}

mhr_model.pt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:352e271a6c42729c68554ceaea0c955e866970160c31e35506d782dc0f7377bc
-size 696110248

model_config.yaml

@@ -1,73 +0,0 @@
-TRAIN:
-  USE_FP16: true
-  FP16_TYPE: bfloat16
-MODEL:
-  IMAGE_SIZE:
-  - 512
-  - 512
-  IMAGE_MEAN:
-  - 0.485
-  - 0.456
-  - 0.406
-  IMAGE_STD:
-  - 0.229
-  - 0.224
-  - 0.225
-  BACKBONE:
-    TYPE: dinov3_vith16plus
-    PRETRAINED_WEIGHTS: ''
-    FROZEN_STAGES: -1
-    DROP_PATH_RATE: 0.1
-  DECODER:
-    TYPE: sam
-    DIM: 1024
-    DEPTH: 6
-    HEADS: 8
-    MLP_DIM: 1024
-    DIM_HEAD: 64
-    LAYER_SCALE_INIT: 0.0
-    DROP_RATE: 0.0
-    ATTN_DROP_RATE: 0.0
-    DROP_PATH_RATE: 0.0
-    FFN_TYPE: origin
-    ENABLE_TWOWAY: false
-    REPEAT_PE: true
-    FROZEN: false
-    CONDITION_TYPE: cliff
-    USE_INTRIN_CENTER: true
-    DO_INTERM_PREDS: true
-    DO_INTERM_SUP: true
-    DO_KEYPOINT_TOKENS: true
-    DO_HAND_DETECT_TOKENS: true
-    KEYPOINT_TOKEN_UPDATE: v2
-    KEYPOINT_TOKEN_UPDATE_COORD_EMB_USE_MLP: true
-    DO_KEYPOINT3D_TOKENS: true
-  PROMPT_ENCODER:
-    ENABLE: true
-    MAX_NUM_CLICKS: 2
-    PROMPT_KEYPOINTS: mhr70
-    FROZEN: false
-    KEYPOINT_SAMPLER:
-      TYPE: v1
-      WORST_RATIO: 0.8
-      KEYBODY_RATIO: 0.8
-      NEGATIVE_RATIO: 0.1
-      DUMMY_RATIO: 0.1
-      DISTANCE_THRESH: 0.0001
-    MASK_EMBED_TYPE: v2
-    MASK_PROMPT: v1
-  PERSON_HEAD:
-    POSE_TYPE: mhr
-    CAMERA_ENABLE: true
-    CAMERA_TYPE: perspective
-    ZERO_POSE_INIT: true
-    ZERO_POSE_INIT_BODY_FACTOR: 1
-  MHR_HEAD:
-    MLP_DEPTH: 2
-    MLP_CHANNEL_DIV_FACTOR: 1
-  CAMERA_HEAD:
-    MLP_DEPTH: 2
-    MLP_CHANNEL_DIV_FACTOR: 1
-    DEFAULT_SCALE_FACTOR_HAND: 10
-  ENABLE_BODY: true
-  ENABLE_HAND: true

motion_native77.yaml

@@ -0,0 +1,31 @@
+# Original native77 final775 Motion with original ViTPose and Body Pose features.
+# For motion-cache, pass motion_native77_e774_s292175.ckpt with --checkpoint
+# and motion_stats_raw.npz with --stats. The codec uses its trained 0.01 std floor.
+network:
+  output_dim: 655
+  max_len: 120
+  cliffcam_dim: 3
+  cam_angvel_dim: 6
+  imgseq_dim: 1024
+  latent_dim: 512
+  num_layers: 12
+  num_heads: 8
+  mlp_ratio: 4.0
+  pred_cam_dim: 3
+  static_conf_dim: 6
+  dropout: 0.1
+  avgbeta: true
+  obs_num_joints: 77
+  sequence_static_slices:
+  - - 456
+    - 584
+  - - 584
+    - 640
+condition_spec:
+  pose_release_id: gem-pose-body-s240000-v1:pose_tokens
+  observation_source: vitpose77-dinov3-huge-metrosim-256x192-v1
+  observation_joint_set: soma77_native_v1
+time_options:
+  reference_fps: 30.0
+  default_fps: 30.0
+  fps_by_data_name: {}

gem_smpl/missing_hmr4d_support/inputs/3DPW/hmr4d_support/test_3dpw_vimo_labels.pt → motion_native77_e774_s292175.ckpt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8fd8a8d1a5a10804257c4fac30c0215c32757aa995d5576b73380eda29f3fe35
-size 466754526
+oid sha256:c7790f4587e518bcea40cb8281c9be2233d37f6caa668835462542760dcf82aa
+size 187134331

scale_mean.pth → motion_stats_raw.npz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f696a0793217f27fcf23fa1fd63e8a34aa74ef64580078698fb9ea7322ce2891
-size 1451
+oid sha256:b7719b634e0fc1f001005605d117d77448c8d4a110923149df58529c68112ed6
+size 21320

onnx/gem_denoiser.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:20aab83c01bbd909a258ad0fa465458eda382c63dc80d56952b2ec952d6e192c
-size 1508810

onnx/gem_denoiser.onnx.data

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:720c2400281b730574bf2a446c33ecd77bf494ac8161b081b09bb122b9fad2f8
-size 199114752

onnx/gem_denoiser_no_imgfeat.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:65c576968bc9548b5c485522930b7985097b70a5e0154e5cd01241d6c806415f
-size 187936462

onnx/gem_denoiser_no_imgfeat.onnx.data

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:5ca91bdf443dc5139335ecef83139572d24ca7765e9cc6dfdd6441ed44bd6278
-size 1156259576

onnx/sam3db_backbone.onnx

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:e03fe953ff9762c3e96b67dcd78e52769e67798cd1f8bd82c4ef5dd7bde48b22
-size 878918

onnx/sam3db_backbone.onnx.data

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:479e7da52d115344a3fd3e7711dae58176f68c723486e80ba2c3efc046f4dbc6
-size 3362053120

sam3d_body.ckpt

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b5a2f9d305dd02626b967aa2e86021fba07065df66ce7a7e00ffb9664f150abf
-size 2109129346

scale_comps.pth

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4c4861384b176fbe3d3bd7480e87ff8161f19553bf6ec1b718bc9b6bbb559cd6
-size 8816