Dataset Hub

The dataset is suitable for feature extraction, audio classification, and voice activity detection tasks. Primarily English language, involving audio deepfake, audio spoofing, voice conversion, deepfake detection, and spoof detection labels, with dataset size ranging from 100 MB to 1 GB.

Biometric Attack Dataset – Different Lighting Conditions Dataset, containing personal and attack photo videos captured under various lighting conditions such as darkroom, daylight, lightroom, and nightlight. Each video in the dataset is about 20 seconds long and is used to learn facial features via deep neural networks and biometric technologies to prevent spoofing attacks.