Datasetpublic

UnFaZeD07/AVE-Dataset

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… See the full description on the dataset page: https://huggingface.co/datasets/UnFaZeD07/AVE-Dataset.

sourceHugging Facemitupdated 9mo agoView on Hugging Face

0likes580downloads

Dataset Card

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",
}