tmddud20/EgoSMPLX

Pretrained Model

This repository provides the pretrained model checkpoint proposed in Egocentric Whole-Body Human Mesh Recovery with Prior-Guided Learning, IEEE ICIP 2026.

The proposed framework performs whole-body human mesh recovery from monocular egocentric images by leveraging an exocentric human mesh recovery foundation model and a diffusion-based pose prior.

For installation, configuration, and evaluation instructions, please refer to the official code repository.

Resources

• Paper (IEEE Xplore)
• arXiv
• Code
• Pseudo-GT Annotations

Citation

If you find our work useful, please cite:

@inproceedings{na2026egocentric,
  author={Na, Soyeon and Noh, Seung Young and Chang, Ju Yong},
  title={Egocentric Whole-Body Human Mesh Recovery with Prior-Guided Learning},
  booktitle={2026 IEEE International Conference on Image Processing (ICIP)},
  year={2026}
}