Fabric Masks and Disguise Presentation Attack Dataset - 9 270 Videos for Face recognition and Computer Vision

Overview The Fabric Masks and Disguise Presentation Attack Dataset is a high-quality face anti-spoofing collection produced by the provider, containing 9,270 videos of 103 people facing the camera while wearing a wide range of fabric and cloth masks. It is purpose-built for presentation attack detection (PAD), face recognition, and liveness detection research, where trained models must reliably tell a real, unmasked human from an impostor using a disguise. The study has placed strong emphasis on variability: the dataset consists of bona fide and attack samples covering different mask types, fabrics, designs, materials, accessories, and lighting environments, giving learning algorithms a realistic foundation for fraud prevention and biometric security tasks. Subject Demographics The dataset covers a balanced and diverse pool of individuals: - Total number of participants: 103 people - Gender split: 50% Male / 50% Female - Ethnicity: Caucasian, African, and Asian participants - Age groups: multiple, captured across both younger and older adults Each video is paired with technical metadata describing the person, the disguise, and the recording environment, supporting supervised classification and detailed analysis of recognition results obtained under masked conditions. Technical Specifications Total videos: 9,270 - Video formats: MP4 and MOV - Resolution: from 1920×1080 up to 3840×2160 (4K) - Clip duration: 3–5 seconds per sample - Backgrounds: 5 unique environments, including white