Face Anti-Spoofing Data

Challenge Dataset URL

Summary

• Introduction
• Dataset Structure
• Reference
• License

Introduction

This dataset is curated for the task of Face Anti-Spoofing, aiming to distinguish between real and spoofed (fake) facial inputs. It includes diverse attack types such as printed photos, screen replays, and 3D masks under various lighting and pose conditions. The dataset supports the development and evaluation of robust liveness detection models for secure face authentication systems.

Dataset Structure

We will release the dataset in three phases. Each phase will provide a .zip file containing face anti-spoofing data with the following composition:

• 70 Real samples: Genuine face images or videos
• 100 Spoof samples: Synthetic inputs including printed photos, screen replays, and other spoofing techniques

Phase 1

Part_1 of the dataset will be released as sample training data(has ground truth file). The remaining data will be reserved for public scoring and evaluation.

Phases 2 & 3

All data will be reserved for public scoring and evaluation in the corresponding phases.

Reference

For more dataset information, please go through the following link, reference dataset.

License

This repository is licensed under the MIT License.