DeepSSM

This is a TensorFlow implementation of DeepSSM as described in the following paper:

Xiaoli Liu, Jianqin Yin, Huaping Liu, Jun Liu. DeepSSM: Deep State-Space Model for 3D Human Motion Prediction[J]. arXiv preprint arXiv:2005.12155, 2020.

Setup

Required python libraries: tensorflow (>=1.0) + opencv + numpy. Tested in ubuntu + GTX 1080Ti with cuda (>=8.0) and cudnn (>=5.0).

Datasets

Human3.6M, 3DPW.

the processed data file will be available at: https://pan.baidu.com/s/1iVsvRC_PUeteY3Oi50teHA （password：123a）

Training

Use the scripts/h36m/DeepSSM_h36m_long_term.sh or scripts/h36m/DeepSSM_h36m_short_term.sh script to train/test the model on Human3.6M dataset for short-term or long-term predictions by the following commands:

cd scripts/h36m
sh DeepSSM_h36m_long_term.sh  # for short-term prediction on Human3.6M
sh DeepSSM_h36m_short_term.sh # for long-term predictions on Human3.6M

You might want to change folders in scripts to train on 3DPW dataset.

Citation

If you use this code for your research, please consider citing:

@article{liu2020deepssm,
  title={DeepSSM: Deep State-Space Model for 3D Human Motion Prediction},
  author={Liu, Xiaoli and Yin, Jianqin and Liu, Huaping and Liu, Jun},
  journal={arXiv preprint arXiv:2005.12155},
  year={2020}
}

Contact

A part of code adopts from PredCNN at https://github.com/xzr12/PredCNN.git.