The implementation of our paper Video-to-Image casting: A flatting method for video analysis.

Run

We use weights pre-trained on imagenet, create a "models/" folder and put these weights into this folder before training.

After downloading the specific dataset (UCF101, DTDB, Something-Something V1), run the following command to train:

python train.py --cfg #File name of each experiment#

Citation

If you find our code useful in your work, please consider using the following citation:

@inproceedings{chen2021video,
  title={Video-to-Image casting: A flatting method for video analysis},
  author={Chen, Xu and Gao, Chenqiang and Yang, Feng and Wang, Xiaohan and Yang, Yi and Han, Yahong},
  booktitle={Proceedings of the 29th ACM International Conference on Multimedia},
  pages={4958--4966},
  year={2021}
}