folder voxceleb (17 files)
filewav/vox2_test_wav.zip 7.70GB
filewav/vox2_dev_wav.zip 231.00GB
filewav/vox1_test_wav.zip 1.07GB
filewav/vox1_dev_wav.zip 32.60GB
filetxt/vox2_test_txt.zip 53.34MB
filetxt/vox2_dev_txt.zip 1.58GB
filetxt/vox1_test_txt.zip 4.57MB
filetxt/vox1_dev_txt.zip 139.08MB
filemeta/vox2_meta.csv 165.24kB
filemeta/vox1_meta.csv 40.78kB
filemeta/veri_test2.txt 2.33MB
filemeta/veri_test.txt 2.34MB
filemeta/list_test_hard2.txt 34.16MB
filemeta/list_test_hard.txt 34.26MB
filemeta/list_test_all2.txt 35.95MB
filemeta/list_test_all.txt 36.05MB
filemeta/iden_split.txt 4.91MB
Type: Dataset
Tags: speaker recognition, speaker identification, speaker verification

Bibtex:
@article{,
title= {voxceleb},
journal= {Proc. Interspeech 2018},
author= {Joon Son Chung and Arsha Nagrani and Andrew Zisserman},
year= {2018},
url= {https://www.robots.ox.ac.uk/~vgg/data/voxceleb},
abstract= {This torrent shares the VoxCeleb1 and VoxCeleb2 datasets. The original dataset creators do not provide access to the dataset anymore. To ensure papers in the field of speaker recognition can be reproduced (many have used VoxCeleb in recent years) the data should be available for academic purposes. 

The audio data is stored as mono-channel, 16000hz, signed 16-bit (little-endian) PCM wav files. This torrent does not include video data.
},
keywords= {speaker recognition, speaker identification, speaker verification},
terms= {The VoxCeleb metadata is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. The audio data is scraped from publicly available YouTube videos. Data is only made available for academic purposes.},
license= {},
superseded= {}
}

Hosted by users:

Send Feedback