WildLife Documentary Dataset
Python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
annotations
frames
subtitles
videos
.gitignore
LICENSE.md
README.md
filelist.txt
video2frames.py

README.md

WildLife Documentary (WLD) Dataset

Introduction

The dataset contains 15 documentary films that are downloaded from YouTube, whose durations vary from 9 minutes to as long as 50 minutes, and the total number of frames is more than 747,000. More than 4000 object tracklets of 65 categories are annotated.

Here is an overview of the dataset. Dataset overview

Content

The dataset are organized as the following structure:

  • videos/: Downloaded raw videos should be extracted here.
  • frames/: Video frames will be generated here.
  • subtitles/: Subtitles of the videos, in srt format. The subtitles are originally auto-generated by YouTube and we correct some obvious mistakes manually.
  • annotations/: Bounding box annotations, in json format. Coordinates are 0-based and the bounding boxes are labeled as [x1, y1, x2, y2]. The videos are fully annotated with the help of object tracking.

Citation

If you use WLD dataset in your research, please consider citing our paper:

@inproceedings{chen2017discover,
  author = {Kai Chen, Hang Song, Chen Change Loy, Dahua Lin},
  title = {Discover and Learn New Objects from Documentaries},
  booktitle = {Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = July,
  year = {2017}
}

Download

  1. Download the raw videos from Google Drive and extract all videos to the folder video/.
  2. run the script video2frames.py (opencv required) to convert all videos into frames.