Truncate-Split-Contrast: A Framework for Learning from Mislabeled Videos [arXiv]

@inproceedings{wang2023truncate,
  title={Truncate-split-contrast: a framework for learning from mislabeled videos},
  author={Wang, Zixiao and Weng, Junwu and Yuan, Chun and Wang, Jue},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={37},
  number={3},
  pages={2751--2758},
  year={2023}
}

Overview

We release the PyTorch code of the Truncate-Split-Contrast.

This code is based on the TSM codebase.

You can train and test by one line bash start.sh

You may need to modify the file opts.py and train.sh to fit your environment.