KNOWN: Body Knowledge and Uncertainty Modeling for Monocular 3D Human Body Reconstruction

Body Knowledge and Uncertainty Modeling for Monocular 3D Human Body Reconstruction
Yufei Zhang, Hanjing Wang, Jeffrey O. Kephart, Qiang Ji
ICCV2023, arXiv

This repository includes the Pytorch demo code and trained models for the above paper.

Installation

conda create -n known python=3.8
conda activate known
pip install -r requirements.txt

Model and Data Download

Please download the trained model and required data from data and store them under the current directory as

./data
├── cam_spin.npy
├── pose_spin.npy
├── shape_spin.npy
├── sigma_all.npy
├── model.pt
├── processed_basicModel_neutral_lbs_10_207_0_v1.0.0.pkl

Demo

KNOWN is the first 3D human body reconstruction model that can quantify both epistemic and aleatoric uncertainty (running in near real-time). We here provide the demo code to generate 3D reconstruction and visualize the 3D vertex prediction uncertainty (epistemic uncertainty).

python demo_img.py --img_path 'path to a testing image'
python demo_live_video.py --video_path 'empty to open a live camera or path to a testing video'

Examples of KNOWN's output are

Citation

If you find our work useful, please consider citing the paper:

@InProceedings{Zhang_2023_ICCV,
    author    = {Zhang, Yufei and Wang, Hanjing and Kephart, Jeffrey O. and Ji, Qiang},
    title     = {Body Knowledge and Uncertainty Modeling for Monocular 3D Human Body Reconstruction},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {9020-9032}
}

If you have questions or encouter any issues when running the code, feel free to open an issue or directly contact me via: [email protected].

References

Part of our code is developed based on SPIN. The SMPL model data is downloaded from SMPL-X model. We thank them for generously sharing their outstanding work.