2D Masks Presentation Attack Detection Dataset

Name: 2D Masks Presentation Attack Detection Dataset
Creator: GTS

Datasets

File

2D Masks Presentation Attack Detection Dataset

Use Case

Face Anti-Spoofing, Liveness Detection, Biometric Security, Presentation Attack Detection, Computer Vision

Description

The 2D Masks Presentation Attack Detection Dataset includes real face videos, printed 2D mask attacks, and cut-out eye mask videos for training and evaluating face anti-spoofing, liveness detection, and biometric security models.

Description:

As facial recognition technology becomes increasingly integrated into mobile devices, banking applications, access control systems, and digital identity platforms, the need for reliable anti-spoofing mechanisms has never been greater. While facial authentication offers convenience and security, it also faces threats from presentation attacks such as printed photos, paper masks, and other spoofing techniques.

Download Dataset

The 2D Masks Presentation Attack Detection Dataset has been developed to support the creation of advanced face liveness detection and anti-spoofing solutions. This dataset contains videos of real individuals, printed 2D mask attacks, and cut-out eye mask attacks captured under diverse environmental conditions. By providing realistic attack scenarios, the dataset enables developers and researchers to train machine learning models capable of distinguishing genuine users from fraudulent attempts.

Dataset Overview

This is a video-based biometric dataset focused on facial presentation attacks.

The dataset contains three main categories:

Real — genuine facial presentations
Mask — printed 2D facial mask attacks
Cut — printed 2D masks with cut-out eye regions

The videos were recorded in different lighting conditions and locations, including indoor and outdoor environments. The sample also includes people wearing accessories such as glasses, caps, hats, and scarves.

Key Features of the Dataset

Real and Spoof Facial Presentations

The dataset includes both genuine facial videos and multiple types of spoof attacks, making it suitable for binary and multi-class classification tasks.

Multiple Attack Types

Researchers can evaluate system performance against different presentation attack methods, including:

Printed 2D face masks
Printed masks with cut-out eye regions
Genuine facial presentations

Video-Based Dataset

Unlike image-only datasets, this collection provides short video sequences that allow models to analyze:

Facial movements
Eye behavior
Texture patterns

Diverse Recording Conditions

Videos are captured in a variety of environments, including:

Indoor locations
Outdoor locations
Different lighting conditions

Dataset Structure

The sample contains 17 participant folders, with 12 videos for each person.

Each participant includes:

Real

Real_1
Real_2
Real_3

Mask

Mask_1
Mask_2
Mask_3

Cut

Cut_1
Cut_2
Cut_3

The accompanying CSV file provides the participant information and references to the corresponding videos, making the dataset easier to organize for machine learning workflows.

Applications of the Dataset

Face Liveness Detection

Develop models that determine whether a facial presentation originates from a live person or a spoofing attack.

Biometric Authentication Systems

Improve facial recognition security in:

Smartphones
Banking applications
Access control systems
Identity verification platforms

Presentation Attack Detection (PAD)

Train systems specifically designed to detect fraudulent biometric presentations.

Computer Vision Research

Support research in:

Facial motion analysis
Temporal feature extraction
Video classification
Deep learning-based security systems

Benefits for Machine Learning and Deep Learning

The dataset offers several advantages for AI model development:

Realistic Attack Scenarios

Includes common attack methods observed in real-world biometric systems.

Video-Based Learning

Supports temporal analysis techniques that improve spoof detection accuracy.

Diverse Environmental Conditions

Helps models generalize across different lighting and background settings.

Multiple Presentation Classes

Enables binary and multi-class classification experiments.

Conclusion

The 2D Masks Presentation Attack Detection Dataset provides a valuable resource for organizations, researchers, and developers working to strengthen biometric security systems. By combining genuine facial presentations with realistic printed mask attacks and cut-out eye spoofing attempts, the dataset enables the development of more accurate and reliable face anti-spoofing solutions.

This dataset is sourced from Kaggle.

Contact Us

FAQ

Question 1. What is the 2D Masks Presentation Attack Detection Dataset?

It is a biometric video dataset containing genuine facial presentations, printed mask attacks, and cut-out eye mask attacks designed for face anti-spoofing and liveness detection research.

Question 2. What types of spoof attacks are included?

The dataset includes:

Printed 2D mask attacks
Printed masks with cut-out eye regions
Genuine facial presentations

Question 3. Is the dataset suitable for liveness detection systems?

Yes. The dataset is specifically designed to support face liveness detection and presentation attack detection applications.

Question 4. Can this dataset be used for deep learning projects?

Absolutely. The dataset is suitable for CNNs, Transformers, LSTMs, and other deep learning architectures used in biometric security applications.

Question 5. Which industries can benefit from this dataset?

Industries including banking, fintech, cybersecurity, access control, identity verification, and biometric authentication can benefit from this dataset.

Let's Discuss your Data collection Requirement With Us

To get a detailed estimation of requirements please reach us.

2D Masks Presentation Attack Detection Dataset