活体检测数据库简介

Database	Year of release	# subjects	# videos	Acquisition camera device	Attack type	Subject race	Subject gender	Subject age
NUAA [1]	2010	15	•24 genuine • 33 spoof	• Web-cam (640 × 480)	• Printed photo	• Asian 100%	• Male 80% • Female 20%	20 to 30 yrs
Idiap REPLAYATTACK [2][3][4]	2012	50	• 200 genuine • 1,000 spoof	• MacBook 13’’   camera (320 × 240)	• Printed photo • Display photo (mobile/HD) • Replayed video (mobile/HD)	• White 76% • Asian 22% • Black 2%	• Male 86% • Female 14%	20 to 40 yrs
CASIA FASD [5]	2012	50	• 150 genuine • 450 spoof	• Low-quality camera (640 × 480) • Normal-quality camera (480 × 640) • Sony NEX-5 camera (1280 × 720)	• Printed photo •Cut photo② • Replayed video (HD)	• Asian 100%	• Male 86% • Female 14%	20 to 35 yrs
MSU MFSD [6]	2014	55①	• 110 genuine • 330 spoof	• MacBook Air 13” camera (640 × 480) • Google Nexus 5 camera (720 × 480)	• Printed photo • Replayed video (mobile/HD)	• White 70% • Asian 28% • Black 2%	• Male 63% • Female 37%	20 to 60 yrs
The Oulu-NPU face anti-spoofing  database	2016	55	• 990 genuine • 3,960 spoof	• Front cameras of six mobile devices（1080×1920） (Samsung Galaxy S6 edge, HTC Desire EYE, MEIZU X5, ASUS Zenfone Selfie, Sony XPERIA C5 Ultra Dual and OPPO N3)	• Two printed photo • Two replayed video (mobile/HD)	• White 4% • Asian 96%	• Male 69% • Female31%	20 to 60 yrs

注：① MSU MFSD中55个人的数据，只有35个人的数据可以公开使用。

②“Cut photo attack”表示将打印图片中眼睛部位剪掉，攻击者用他的照片盖住他的脸，并且可以在洞里眨眼睛。

CASIA FASD 3D-MAD database Print attack

MSU-MFSD

anti-spoofing， face-detect等数据集

活体检测：

Reference

1 人脸相关数据库

2 Face 研究可用数据集