Long-term Action Forecasting Using Multi-headed Attention-based Variational Recurrent Neural Networks (MAVAP)

Pytorch implementation of MAVAP, a probablistic neural network archicture for action forecasting from videos. This piece of work was presented at the 3rd Workshop and Competition on Affective Behavior Analysis in-the-wild (ABAW), which was hosted at the Computer Vision and Pattern Recognition conference in 2022.

@inproceedings{loh2022long,
  title={Long-term Action Forecasting Using Multi-headed Attention-based Variational Recurrent Neural Networks},
  author={Loh, Siyuan Brandon and Roy, Debaditya and Fernando, Basura},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={2419--2427},
  year={2022}
}