HKBU MARs Database - The HKBU 3D Mask Attack with Real World Variations Database

As the increasing deployment of face recognition in a variety of applications, its security concern becomes increasingly important. 3D mask attack becomes new challenge and attracts more research interests. The HKBU 3D Mask Attack with Real World Variations Database (HKBU-MARs) is a 3D mask attack database with more variations to simulate the real world application scenario.

Overview

Version 1+

HKBU-MARs V1+ is the dataset used in our ECCV'18 paper (S. Q. Liu et al., ECCV'18). Based on HKBU-MARs V1, the version 1+ adds 4 customized REAL-f masks and 4 subjects (subject index 13-16). The HKBU-MARs V1+ follows the same organization of HKBU-MARs V1. For more details please refer to our paper and HKBU-MARs V1.

Version 1

HKBU-MARs V1 is the SUP dataset introduced in our ECCV'16 paper (S. Q. Liu et al., ECCV'16). It contains 120 videos from 8 subjects and masks. 2 types of 3D masks are included: 6 from Thatsmyface and 2 from REAL-f. Each subject has 10 genuine samples and 5 masked samples. All videos are recorded through Logeitech C920 web-camera in the resolution of 1280*720 under room lighting condition. The recordings are done with frontal-view and neutral expression. For more details please refer (S. Q. Liu et al.,ECCV'16).

Figure 1. Example images of genuine and spoof faces HKBU-MARs V1. Bottom row shows the ThatsMyFace (Left) and Real-F (Right) masks included in our database.

Organization: The HKBU-MARs V1 dataset is orgnized in 3 different sessions for all subjects. For each subject in a single session, it has 5 videos of 300 frames for each. The first two sessions are the real access samples. In the third session, 3D mask attacks are captured by a single operator (attacker).

Version 2

HKBU-MARs V2 is the database proposed in our CVPRW'16 paper (S. Q. Liu et al., CVPRW'16), which covers the variations in real world applicaton. It contains 1008 videos from 12 subjects and masks. 2 types of 3D masks are included: 6 from Thatsmyface and 6 from REAL-f. 7 cameras from the stationary and mobile devices and 6 lighting settings that cover typical illumination conditions are also included. For more details please refer (S. Q. Liu et al.,CVPRW'16)

Camera Variations:

Lighting Variations:

This version is released now. Contact Siqi Liu for further details.

DownLoad

Agreement

Siqi Liu (siqiliu@comp.hkbu.edu.hk)

References:

S. Q. Liu, X. Lan, and P. C. YUEN, "Multi-Channel Remote Photoplethysmography Correspondence Feature for 3D Mask Face Presentation Attack Detection," TIFS'21
S. Q. Liu, X. Lan, and P. C. YUEN, "Temporal Similarity Analysis of Remote Photoplethysmography for Fast 3D Mask Face Presentation Attack Detection," WACV'20
rPPG Correspondence Feature for 3D Mask Face Presentation Attack Detection, ECCV'18, (HKBU-MARs V1+)
S. Q. Liu, X. Lan, and P. C. YUEN
3D Mask Face Anti-spoofing with Remote Photoplethysmography, ECCV'16, (HKBU-MARs V1)
S. Q. Liu, P. C. YUEN, S. Zhang, and G. Zhao
A 3D Mask Face Anti-spoofing Database with RealWorld Variations, CVPRW'16, (HKBU-MARs V2)
S. Q. Liu, B. Yang, P. C. YUEN, and G. Zhao
Multi-adversarial Discriminative Deep Domain Generalization for Face Presentation Attack Detection, CVPR'19
R. Shao, X. Lan, J. Li, and P. C. YUEN
Joint Discriminative Learning of Deep Dynamic Textures for 3D Mask Face Anti-Spooﬁng, TIFS'19
R. Shao, X. Lan, and P. C. YUEN