Skip to content

Latest commit

 

History

History
48 lines (30 loc) · 2.2 KB

README.md

File metadata and controls

48 lines (30 loc) · 2.2 KB
Raw

Moment Detection in Long Tutorial Videos

This repository contains the code for LongMoment-DETR, a method designed for moment detection in long tutorial videos. Additionally, it houses two datasets: Behance Moment Detection (BMD) and YouTube Chapters (YTC), tailored for the same purpose. Our ICCV 2023 paper (supplement) contains more details about this work.

Code and Data

Please visit longmoment-detr for the code and data.

Table of Contents

BMD Dataset

The BMD dataset is curated for moment detection in long tutorial videos.

  • Location: bmd_data/
  • Details: For a comprehensive description of the dataset, its structure, and annotation procedure, please refer to the bmd_data README.

YTC Dataset

The YTC dataset provides annotations for YouTube chapters in long tutorial videos, offering structured segmentations derived directly from YouTube.

  • Location: ytc_data/
  • Details: For more information about the dataset, its structure, and content, please check the ytc_data README.

Citation

If you use this work in your research, please cite our ICCV 2023 paper (supplement) as follows:

@inproceedings{croitoru2023moment,
  title={Moment detection in long tutorial videos},
  author={Croitoru, Ioana and Bogolin, Simion-Vlad and Albanie, Samuel and Liu, Yang and Wang, Zhaowen and Yoon, Seunghyun and Dernoncourt, Franck and Jin, Hailin and Bui, Trung},
  booktitle={Proceedings of the IEEE/CVF international conference on computer vision},
  pages={2594--2604},
  year={2023}
}