Code and data will be released soon (by July).
Here are the Google Drive links to the dataset, including features and annotations. Due to the high volume, I am looking for another way to release the features for YouTube videos (Untrimmed Kinetics).
[FineAction] [THUMOS14] [ANET13]
@article{hyun2024exploring,
title={Exploring Scalability of Self-Training for Open-Vocabulary Temporal Action Localization},
author={Jeongseok Hyun and Su Ho Han and Hyolim Kang and Joon-Young Lee and Seon Joo Kim},
journal={arXiv preprint arXiv:2407.07024},
year={2024}
}