This repository contains the code for TEMMA.
The main.py gives an example of how to use the CNN-TE as well as the CNN-TEMMA models.
For the multi-modal CNN-TEMMA, please note that the input multi-stream features need to be concatenated first as in the example of the main.py
if you find this repository helpful, please cite our publication:
@ARTICLE{9257201,
author={H. {Chen} and D. {Jiang} and H. {Sahli}},
journal={IEEE Transactions on Multimedia},
title={Transformer Encoder with Multi-modal Multi-head Attention for Continuous Affect Recognition},
year={2020},
doi={10.1109/TMM.2020.3037496}}
Contact person: Haifeng Chen, Email: [email protected]
- The codes of basemodel_1D is borrowed from TCN.
- The structure of the codebase is borrowed from The Annotated Transformer