Holistic Video Understanding Mini Dataset

This repository contains the Holistic Video Understanding Mini Dataset. For more information regarding the dataset please check our paper.

⚠️Note! The provided labels for this version are machine generated. The complete version of HVU with clean and human-verified annotations will be released here.

If you use our dataset, please cite our paper.

@article{hvu2019,
  title={Large Scale Holistic Video Understanding},
  author={Diba, Ali and Fayyaz, Mohsen and Sharma, Vivek and Paluri, Manohar and Gall, J{\"u}rgen and Stiefelhagen, Rainer and Van Gool, Luc},
  journal={arXiv preprint arXiv:1904.11451},
  year={2019}
}

Table of Contents

• Downloading
  • ICCV 2019 Release
• Usage

Downloading

ICCV 2019 Release

To download our dataset release for the Holistic Video Understanding workshop in ICCV 2019, use the following:

git clone -b iccv2019 --depth 1 [email protected]:holistic-video-understanding/Mini-HVU.git

Usage

We store the video IDs from Youtube and their annotations in CSV format.

Dataset Downloader

Check HVU downloader instruction for an easy way to download HVU.