Hugging Face's logo

walterzhu
/
MotionBERT

3D Human Pose Estimation
Skeleton-based Action Recognition
Mesh Recovery
Model card Files Community
2
walterzhu commited on
Commit
370a919
•
1 Parent(s): 13ac6c1

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -7
README.md CHANGED
@@ -76,13 +76,13 @@ E = MotionBERT.get_representation(x)
76
 
77
  | Model | Download Link | Config | Performance |
78
  | ------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ---------------- |
79
- | MotionBERT (162MB) | [OneDrive](https://1drv.ms/f/s!AvAdh0LSjEOlgS425shtVi9e5reN?e=6UeBa2) | [pretrain/MB_pretrain.yaml](configs/pretrain/MB_pretrain.yaml) | - |
80
- | MotionBERT-Lite (61MB) | [OneDrive](https://1drv.ms/f/s!AvAdh0LSjEOlgS27Ydcbpxlkl0ng?e=rq2Btn) | [pretrain/MB_lite.yaml](configs/pretrain/MB_lite.yaml) | - |
81
- | 3D Pose (H36M-SH, scratch) | [OneDrive](https://1drv.ms/f/s!AvAdh0LSjEOlgSvNejMQ0OHxMGZC?e=KcwBk1) | [pose3d/MB_train_h36m.yaml](configs/pose3d/MB_train_h36m.yaml) | 39.2mm (MPJPE) |
82
- | 3D Pose (H36M-SH, ft) | [OneDrive](https://1drv.ms/f/s!AvAdh0LSjEOlgSoTqtyR5Zsgi8_Z?e=rn4VJf) | [pose3d/MB_ft_h36m.yaml](configs/pose3d/MB_ft_h36m.yaml) | 37.2mm (MPJPE) |
83
- | Action Recognition (x-sub, ft) | [OneDrive](https://1drv.ms/f/s!AvAdh0LSjEOlgTX23yT_NO7RiZz-?e=nX6w2j) | [action/MB_ft_NTU60_xsub.yaml](configs/action/MB_ft_NTU60_xsub.yaml) | 97.2% (Top1 Acc) |
84
- | Action Recognition (x-view, ft) | [OneDrive](https://1drv.ms/f/s!AvAdh0LSjEOlgTaNiXw2Nal-g37M?e=lSkE4T) | [action/MB_ft_NTU60_xview.yaml](configs/action/MB_ft_NTU60_xview.yaml) | 93.0% (Top1 Acc) |
85
- | Mesh (with 3DPW, ft) | [OneDrive](https://1drv.ms/f/s!AvAdh0LSjEOlgTmgYNslCDWMNQi9?e=WjcB1F) | [mesh/MB_ft_pw3d.yaml](configs/mesh/MB_ft_pw3d.yaml) | 88.1mm (MPVE) |
86
 
87
  In most use cases (especially with finetuning), `MotionBERT-Lite` gives a similar performance with lower computation overhead.
88
 
 
76
 
77
  | Model | Download Link | Config | Performance |
78
  | ------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | ---------------- |
79
+ | MotionBERT (162MB) | [HuggingFace](https://huggingface.co/walterzhu/MotionBERT/blob/main/checkpoint/pretrain/MB_release/latest_epoch.bin) | [pretrain/MB_pretrain.yaml](configs/pretrain/MB_pretrain.yaml) | - |
80
+ | MotionBERT-Lite (61MB) | [HuggingFace](https://huggingface.co/walterzhu/MotionBERT/blob/main/checkpoint/pretrain/MB_lite/latest_epoch.bin) | [pretrain/MB_lite.yaml](configs/pretrain/MB_lite.yaml) | - |
81
+ | 3D Pose (H36M-SH, scratch) | [HuggingFace](https://huggingface.co/walterzhu/MotionBERT/blob/main/checkpoint/pose3d/MB_train_h36m/best_epoch.bin) | [pose3d/MB_train_h36m.yaml](configs/pose3d/MB_train_h36m.yaml) | 39.2mm (MPJPE) |
82
+ | 3D Pose (H36M-SH, ft) | [HuggingFace](https://huggingface.co/walterzhu/MotionBERT/blob/main/checkpoint/pose3d/FT_MB_release_MB_ft_h36m/best_epoch.bin) | [pose3d/MB_ft_h36m.yaml](configs/pose3d/MB_ft_h36m.yaml) | 37.2mm (MPJPE) |
83
+ | Action Recognition (x-sub, ft) | [HuggingFace](https://huggingface.co/walterzhu/MotionBERT/blob/main/checkpoint/action/FT_MB_release_MB_ft_NTU60_xsub/best_epoch.bin) | [action/MB_ft_NTU60_xsub.yaml](configs/action/MB_ft_NTU60_xsub.yaml) | 97.2% (Top1 Acc) |
84
+ | Action Recognition (x-view, ft) | [HuggingFace](https://huggingface.co/walterzhu/MotionBERT/blob/main/checkpoint/action/FT_MB_release_MB_ft_NTU60_xview/best_epoch.bin) | [action/MB_ft_NTU60_xview.yaml](configs/action/MB_ft_NTU60_xview.yaml) | 93.0% (Top1 Acc) |
85
+ | Mesh (with 3DPW, ft) | [HuggingFace](https://huggingface.co/walterzhu/MotionBERT/blob/main/checkpoint/mesh/FT_MB_release_MB_ft_pw3d/best_epoch.bin) | [mesh/MB_ft_pw3d.yaml](configs/mesh/MB_ft_pw3d.yaml) | 88.1mm (MPVE) |
86
 
87
  In most use cases (especially with finetuning), `MotionBERT-Lite` gives a similar performance with lower computation overhead.
88