Correlative Channel-Aware Fusion for Multi-View Time Series Classification (C2AF)

This repository is for our AAAI'21 paper:

Correlative Channel-Aware Fusion for Multi-View Time Series Classification PDF
Yue Bai, Lichen Wang, Zhiqiang Tao, Sheng Li, and Yun Fu

This paper proposes to use a learnable fusion strategy to enhance the multi-view time series classification performance. The proposed correlative channel-aware fusion module can be simply realized by a convolutional filter yet effective for final fusion performance. We originally implement it using Tensorflow but it can be easily revised for Pytorch platform.

Usage

Please directly check the demo.py file which contains the whole pipeline to train the model.

Extracted feature

In our work, we use a newly proposed multi-view action dataset (EV-Action). Here, we provide our extracted feature for usage, where RGB, depth, and skeleton features are aligned in temporal dimension and can be used directly to train a model. Please check the google drive link as below.

Extracted EV-Action feature

Reference

Please cite this in your publication if our code or feature helps your research. Should you have any questions, welcome to reach out to Yue Bai ([email protected]).

@inproceedings{bai2021correlative,
  title={Correlative channel-aware fusion for multi-view time series classification},
  author={Bai, Yue and Wang, Lichen and Tao, Zhiqiang and Li, Sheng and Fu, Yun},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={35},
  number={8},
  pages={6714--6722},
  year={2021}
}