UniDataPro/fabric-masks-dataset

Face Mask Detection Dataset

The dataset consists of 9,270 videos featuring 103 people wearing various fabric face masks, captured under diverse conditions. It is designed for research in presentation attack detection (PAD), focusing on challenging facial recognition systems and enhancing fraud prevention mechanisms.

Researchers can utilize this data for developing advanced mask detection and face recognition algorithms. - Get the data

The videos showcase a wide variety of mask types and designs, worn by individuals from different age groups and ethnicities. The study has placed strong emphasis on variability, including factors like worn masks, the use of glasses or wigs, and changing light conditions and backgrounds.

💵 Buy the Dataset: This is a limited preview of the data. To access the full dataset, please contact us at https://unidata.pro to discuss your requirements and pricing options.

Metadata for the dataset

Variables in .csv files:

• person_id: The unique identifier for the genuine participant.
• sample_id: The unique identifier for each video sample.
• gender: Gender of the participant (Male/Female).
• age: The approximate age of the participant.
• class: The type of sample (bona_fide for real, attack for presentation attack).
• impostor_id: For attack samples, the ID of the impostor wearing the mask (e.g., IMP_A); is none for bona fide samples.
• mask_id: For attack samples, the ID of the specific fabric mask used (e.g., MASK_1_A); is none for bona fide samples.
• glasses: Indicates if glasses are present (e.g., rimless, none).
• wig: Indicates if a wig is present (e.g., short_blond, short_dark, none).
• camera: The recording device used (e.g., Iphone, Samsung).
• light_condition: The lighting environment (e.g., natural, artificial, natural_and_artificial, dim_light).
• background: The backdrop of the recording (e.g., white_wall, office_bookshelves, office_brick_wall, window).

This detailed metadata provides a robust foundation for achieving higher detection accuracy, advancing liveness detection methods.

Frequently Asked Questions

What recording quality does the fabric masks dataset provide?

The fabric masks dataset contains high-definition video recordings captured at resolutions ranging from 1920 × 1080 (Full HD) up to 3840 × 2160 (4K).

Who can benefit from this fabric masks dataset?

This masks dataset is valuable for biometric security companies, AI researchers, computer vision engineers, identity verification providers, financial institutions, cybersecurity teams, mobile authentication developers, universities, and organizations building secure facial recognition systems.

What demographic diversity is included in the face anti-spoofing dataset?

The dataset contains recordings from 103 participants with balanced gender representation and includes individuals from Caucasian, African, and Asian backgrounds.

🌐 UniData provides high-quality datasets, content moderation, data collection and annotation for your AI/ML projects