Zeusee Face Anti spoofing

Zeusee cooperates with live facial detection

When developing facial recognition applications on mobile terminals, users often use remake photos and videos to deceive the facial recognition system. Therefore, live detection is a very important part of face recognition and face verification. Currently, there are few open source live detection codes, but the implementation of live detection is relatively simple, so we have made a simple implementation of it and open source all the codes.

renew

• Live detection based on HyperLandmark

• Added IntraFace-based live detection (2018.03.30)

Monocular-coordinated live detection

Due to the limitations of mobile sensors, in the case of a wide range of single cameras, the implementation of non-coordinated facial liveness detection is relatively complex. The data required is also relatively high. Cooperating with sexual live detection only requires users to simply lower their heads and shake their heads, which can greatly reduce the cost of fraud attacks.

Basic design ideas reference

Tips: Some of the processes above are not fully implemented

Use of CPP code

C++ code only provides pose detection

# include " AliveDetector.h "
int main ()
{
    AliveDetector *aliveDetector = new AliveDetector (detector_path);
    while ( /* <read Frame> */ ){
        int state = aliveDetector-> detect ( Frame);
        // 检测一帧返回人脸状态
        // 检测不出人脸或人脸数超过1    -1
        // 正常                        0
        // 摇头                        1
        // 低头                        2
        // 抬头                        3
        }
    return 0 ;
}

Android Demo

1. Provides simple demo project 2. NDK is version 16

Contact and communicate:

Regarding monocular non-cooperated live detection, we are conducting data collection and algorithm development. Technical exchanges and donation data can be added to the next group.

• Communication group 724142079
• Live detection technology blog