Improve model card for SKEL-CF: add metadata, links, description, usage, and visuals (#1)
Browse files- Improve model card for SKEL-CF: add metadata, links, description, usage, and visuals (c10c15b162d27661538b18ea63d821f5cfb58bc2)
Co-authored-by: Niels Rogge <nielsr@users.noreply.huggingface.co>
README.md
CHANGED
|
@@ -1,15 +1,100 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
| 4 |
|
| 5 |
-
#
|
| 6 |
|
| 7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
### Model Sources
|
| 10 |
|
| 11 |
<!-- Provide the basic links for the model. -->
|
| 12 |
|
| 13 |
-
-
|
| 14 |
-
-
|
| 15 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
pipeline_tag: image-to-3d
|
| 4 |
+
tags:
|
| 5 |
+
- human-mesh-recovery
|
| 6 |
---
|
| 7 |
|
| 8 |
+
# SKEL-CF: Coarse-to-Fine Biomechanical Skeleton and Surface Mesh Recovery
|
| 9 |
|
| 10 |
+
TL;DR: SKEL-CF is a coarse-to-fine transformer framework for estimating anatomically accurate SKEL parameters from 3D human data. By converting 4DHuman to SKEL-aligned 4DHuman-SKEL and incorporating camera modeling, it addresses data scarcity and depth/scale ambiguities. SKEL-CF outperforms prior SKEL-based methods on MOYO (85.0 MPJPE / 51.4 PA-MPJPE), offering a scalable, biomechanically faithful solution for human motion analysis.
|
| 11 |
+
|
| 12 |
+
<p align="center">
|
| 13 |
+
<img src="https://raw.githubusercontent.com/Pokerman8/SKEL-CF/main/static/images/teaser.jpg" alt="SKEL-CF Teaser Image">
|
| 14 |
+
</p>
|
| 15 |
|
| 16 |
### Model Sources
|
| 17 |
|
| 18 |
<!-- Provide the basic links for the model. -->
|
| 19 |
|
| 20 |
+
- **Repository:** [Intellindust-AI-Lab/SKEL-CF](https://github.com/Intellindust-AI-Lab/SKEL-CF)
|
| 21 |
+
- **Paper (Hugging Face):** [SKEL-CF: Coarse-to-Fine Biomechanical Skeleton and Surface Mesh Recovery](https://huggingface.co/papers/2511.20157)
|
| 22 |
+
- **Paper (arXiv):** [SKEL-CF: Coarse-to-Fine Biomechanical Skeleton and Surface Mesh Recovery](http://arxiv.org/abs/2511.20157)
|
| 23 |
+
- **Project Page:** [SKEL-CF Project Page](https://pokerman8.github.io/SKEL-CF/)
|
| 24 |
+
- **Demo:** [Github Demo](https://github.com/Intellindust-AI-Lab/SKEL-CF/blob/main/run_hsmr_test.py)
|
| 25 |
+
|
| 26 |
+
---
|
| 27 |
+
|
| 28 |
+
## ποΈ Updates
|
| 29 |
+
- [x] **\[2025.11.26\]** Release SKEL-CF.
|
| 30 |
+
- [x] **\[2025.11.27\]** Release checkpoints and labels on [Hugging Face](https://huggingface.co/Intellindust/SKEL_CF_vitpose_H).
|
| 31 |
+
|
| 32 |
+
## π§ Table of Content
|
| 33 |
+
* [1. βοΈ Setup](#%EF%B8%8F-setup)
|
| 34 |
+
* [2. π Demo & Quick Start](#-demo--quick-start)
|
| 35 |
+
* [3. π§± Reproducibility](#-reproducibility)
|
| 36 |
+
* [4. π Visual Results](#-visual-results)
|
| 37 |
+
* [5. π Citation](#-citation)
|
| 38 |
+
* [6. π Acknowledgement](#-acknowledgement)
|
| 39 |
+
* [7. π Star History](#-star-history)
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
## βοΈ Setup
|
| 43 |
+
|
| 44 |
+
1. [π Environment Setup](./docs/SETUP.md#environment-setup)
|
| 45 |
+
2. [π¦ Data Preparation](./docs/SETUP.md#data-preparation)
|
| 46 |
+
|
| 47 |
+
|
| 48 |
+
## π Quick Start
|
| 49 |
+
Quick start with images:
|
| 50 |
+
|
| 51 |
+
```shell
|
| 52 |
+
bash vis/run_demo.sh
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
Quick start with videos:
|
| 56 |
+
|
| 57 |
+
```shell
|
| 58 |
+
bash vis/run_video.sh
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
## π§± Reproducibility
|
| 62 |
+
|
| 63 |
+
For reproducing the results in the paper, please refer to [`docs/EVAL.md`](./docs/EVAL.md) and [`docs/TRAIN.md`](./docs/TRAIN.md).
|
| 64 |
+
|
| 65 |
+
## π Visual Results
|
| 66 |
+
|
| 67 |
+
### Per-Layer Refinement
|
| 68 |
+

|
| 69 |
+

|
| 70 |
+
### Sports Video
|
| 71 |
+
<a href="https://raw.githubusercontent.com/Pokerman8/SKEL-CF/main/static/videos/badminton01.mp4">
|
| 72 |
+
<img src="https://img.shields.io/badge/βΆοΈ-Badminton%20Video-FF6B6B?style=for-the-badge&logo=video" alt="Badminton Video" />
|
| 73 |
+
</a>
|
| 74 |
+
|
| 75 |
+
|
| 76 |
+
<a href="https://raw.githubusercontent.com/Pokerman8/SKEL-CF/main/static/videos/skate01.mp4">
|
| 77 |
+
<img src="https://img.shields.io/badge/βΆοΈ-Skate%20Video-4ECDC4?style=for-the-badge&logo=video" alt="Skate Video" />
|
| 78 |
+
</a>
|
| 79 |
+
|
| 80 |
+
> π‘ **Tip**: Click the buttons above to watch videos, or visit our [project page](https://pokerman8.github.io/SKEL-CF/) for more visual results.
|
| 81 |
+
|
| 82 |
+
|
| 83 |
+
## π Citation
|
| 84 |
+
If you use `SKEL-CF` or its methods in your work, please cite the following BibTeX entries:
|
| 85 |
+
|
| 86 |
+
```latex
|
| 87 |
+
@article{li2025skelcf,
|
| 88 |
+
title={SKEL-CF: Coarse-to-Fine Biomechanical Skeleton and Surface Mesh Recovery},
|
| 89 |
+
author={Li, Da and Jin, Jiping and Yu, Xuanlong and Cun, Xiaodong and Chen, Kai and Fan, Rui and Kong, Jiangang and Shen, Xi},
|
| 90 |
+
journal={arXiv},
|
| 91 |
+
year={2025}
|
| 92 |
+
}
|
| 93 |
+
```
|
| 94 |
+
## π Acknowledgement
|
| 95 |
+
|
| 96 |
+
Parts of the code are adapted from the following repos: [SKEL](https://github.com/MarilynKeller/SKEL), [CameraHMR](https://github.com/pixelite1201/CameraHMR), [HSMR](https://github.com/IsshikiHugh/HSMR), [ViTPose](https://github.com/ViTAE-Transformer/ViTPose), [Detectron2](https://github.com/facebookresearch/detectron2)
|
| 97 |
+
|
| 98 |
+
## π Star History
|
| 99 |
+
|
| 100 |
+
[](https://www.star-history.com/#Intellindust-AI-Lab/SKEL-CF&type=date&legend=top-left)
|