metadata
license: mit
size_categories:
- 1K<n<10K
tags:
- Audio-Visual-Event-Localization
AVE Dataset
The original AVE dataset ported from the GitHub
Notes
- One video may contain different audio-visual events, so the total number of videos is not 4143.
- annotations.txt: annotations of the AVE dataset. For each sample, you can find its event category, YouTube ID, Quality (all good, meaning that it contains an AVE), start time of an audio-visual event, and the end time of an audio-visual event.
- train/val/test-Set.txt: training/validation/testing set used in the original paper.
Citation
If you use the AVE dataset, please consider citing the original paper:
@inproceedings{TianECCV2018,
title={Audio-Visual Event Localization in Unconstrained Videos},
author={Tian, Yapeng and Shi, Jing and Li, Bochen and Duan, Zhiyao and Xu, Chenliang},
editor="Ferrari, Vittorio and Hebert, Martial and Sminchisescu, Cristian and Weiss, Yair",
booktitle="Computer Vision -- ECCV 2018",
year="2018",
publisher="Springer",
}