A very simple baseline to estimate 2D & 3D SMPL-compatible keypoints from a single color image.

The model file is only 51.2 MB and runs at 70fps on a 2080Ti.

1. Download the pre-trained model here into model/.
2. Install the dependencies.
3. Run python example.py.

The input image should already be cropped to a 4:3 portrait with the subject in the center.

We will not release the training code.

The model is trained using the following datasets jointly:

• COCO
• MPII2D
• MPII3D
• HUMBI
• HM36M
• MTC
• SPIN

Please also check the license of the listed datasets.

If you find it helpful, please consider citing our related paper:

@InProceedings{Zhou_2021_CVPR,
    author    = {Zhou, Yuxiao and Habermann, Marc and Habibie, Ikhsanul and Tewari, Ayush and Theobalt, Christian and Xu, Feng},
    title     = {Monocular Real-Time Full Body Capture With Inter-Part Correlations},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2021},
    pages     = {4811-4822}
}