DSTD-GCN

This is the official pytorch implementation of DSTD-GCN.

This paper is accepted by IEEE Transactions on Neural Networks and Learning Systems.

Overview

Human motion prediction is challenging due to the complex spatiotemporal feature modeling. Among all methods, graph convolution networks (GCNs) are extensively utilized because of their superiority in explicit connection modeling. Within a GCN, the graph correlation adjacency matrix drives feature aggregation, and thus, is the key to extracting predictive motion features. State-of-the-art methods decompose the spatiotemporal correlation into spatial correlations for each frame and temporal correlations for each joint. Directly parameterizing these correlations introduces redundant parameters to represent common relations shared by all frames and all joints. Besides, the spatiotemporal graph adjacency matrix is the same for different motion samples, and thus, cannot reflect samplewise correspondence variances. To overcome these two bottlenecks, we propose dynamic spatiotemporal decompose GC (DSTD-GC), which only takes 28.6% parameters of the state-of-the-art GC. The key of DSTD-GC is constrained dynamic correlation modeling, which explicitly parameterizes the common static constraints as a spatial/temporal vanilla adjacency matrix shared by all frames/joints and dynamically extracts correspondence variances for each frame/joint with an adjustment modeling function. For each sample, the common constrained adjacency matrices are fixed to represent generic motion patterns, while the extracted variances complete the matrices with specific pattern adjustments. Meanwhile, we mathematically reformulate GCs on spatiotemporal graphs into a unified form and find that DSTD-GC relaxes certain constraints of other GC, which contributes to a better representation capability. Moreover, by combining DSTD-GC with prior knowledge like body connection and temporal context, we propose a powerful spatiotemporal GCN called DSTD-GCN. On the Human3.6M, Carnegie Mellon University (3dpw) Mocap, and 3D Poses in the Wild (3DPW) datasets, DSTD-GCN outperforms state-of-the-art methods by 3.9%–8.7% in prediction accuracy with 55.0%–96.9% fewer parameters.

Setup

Training & Evaluation

For all the training and evaluation codes, we provide scripts below. Besides, you can also find running scripts under scripts/.

We also provide pretrained model checkpoints, which can be downloaded from Google Drive or Baidu Netdisk (extraction code: dstd). Please download the checkpoints and place it under milestone/.

Cutomization

You can also develop your own method or set up your own dataset in this project.

TODOS

• [x] Model checkpoints
• [x] Customization codes

Citation

If you find this project contributes to your work, please cite the following paper.

@article{fu2023dstdgcn, author={Fu, Jiajun and Yang, Fuxing and Dang, Yonghao and Liu, Xiaoli and Yin, Jianqin}, journal={IEEE Transactions on Neural Networks and Learning Systems}, title={Learning Constrained Dynamic Correlations in Spatiotemporal Graphs for Motion Prediction}, year={2024}, doi={10.1109/TNNLS.2023.3277476}, volume={35}, number={10}, pages={14273-14287} }

Acknowledgements

Some of the codes are from STS-GCN, LearnTrajDep and MSR-GCN. Besides, this project is developed based on the template. We really thank the authors for their wonderful works.