This is official implementation for the paper

Generating Smooth Pose Sequences for Diverse Human Motion Prediction. In ICCV 21.

Wei Mao, Miaomiao Liu, Mathieu Salzmann.

[paper] [talk]

• Python >= 3.8
• PyTorch >= 1.8
• Tensorboard

tested on pytorch == 1.8.1

• We follow the data preprocessing steps (DATASETS.md) inside the VideoPose3D repo.
• Given the processed dataset, we further compute the multi-modal future for each motion sequence. All data needed can be downloaded from Google Drive and place all the dataset in data folder inside the root of this repo.

• We provide 4 YAML configs inside motion_pred/cfg: [dataset].yml and [dataset]_nf.yml for training generator and normalizing flow respectively. These configs correspond to pretrained models inside results.
• The training and evaluation command is included in run.sh file.

If you use our code, please cite our work

@inproceedings{mao2021generating,
  title={Generating Smooth Pose Sequences for Diverse Human Motion Prediction},
  author={Mao, Wei and Liu, Miaomiao and Salzmann, Mathieu},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={13309--13318},
  year={2021}
}

The overall code framework (dataloading, training, testing etc.) is adapted from DLow.

MIT