silent-face-anti-spoofing

Purpose

The silent face anti-spoofing detection model is used to determine if the face in an image is real or fake.

It is designed to prevent people from tricking facial identification systems, such as those used for unlocking phones or accessing secure locations. This is achieved through a process called "liveness" or "anti spoofing" which judges whether the face presented is genuine or not.

The face presented by other media can be defined as a fake: photo prints of faces, faces on phone screens, silicone mask, 3D human image, etc.

This model outputs three concepts:

• fake2d
• fake3d
• real

Architecture

This model adopts a silent living detection method based on the auxiliary supervision of Fourier spectrum which can reflect the difference of true and false faces in frequency domain to a certain extent. The overall architecture is shown in the following figure:

By using our self-developed model pruning method, the FLOPs of MobileFaceNet is reduced from 0.224G to 0.081G, and the performance of the model is significantly improved (the amount of calculation and parameters is reduced) with little loss of precision.

Limitations

• Images must be collected by camera, otherwise it does not conform to the normal scene usage specification, and the algorithm effect cannot be guaranteed.
• As RGB silent living body detection depends on camera model and scene, the experience could vary.
• Face should appear completely in the view. Detection will be affected when the face is not within 30 degrees of the normal face recognition scene, horizontally or vertically.

Sources

Silent liveness detection (Silent-Face-Anti-Spoofing) on GitHub