Real-Time Face Liveness Detection and Face Anti-spoofing Using Deep Learning
Ruchi Zawar, Vrishali A. Chakkarwar · Advances in intelligent systems research/Advances in Intelligent Systems Research · 2023
Face recognition biometrics is now widely employed, thanks to the rapid development of computer vision technology.However, since the facial recognition system cannot tell whether a face image is real or not, it is open to impersonation attempts.A face recognition system should be able to recognize not just people's faces but also spoofing attempts using printed images, videos, or 3D masks.Examining facial liveness, such as landmark detection, eye blinking, and lip movement is a genuine strategy to avoid spoofing.However, when it comes to video-based replay attacks, this strategy is not sufficient.As a result, this study provides a face liveness detection approach that is integrated with a CNN (Convolutional Neural Network) classifier.The landmark detector module identifies the facial landmarks, the eye module analyses eye liveness, and the CNN classifier module makes up the anti-spoofing approach.We created an anti-spoofing model based on MobileNetV2, which was altered and retrained effectively using the LCC FASD dataset, which is freely available for this purpose.In an effort to get rapid inference time with satisfactory precision, a MobileNetV2's transfer learning is used as the classifier.We subsequently merged these landmark detection, eye liveness detection, and anti-spoofing modules and used the combined result to create a simple facial anti-spoofing and liveness detection application.The test results reveal that the built module can distinguish a variety of facial spoof attacks and has a high level of accuracy of 98%.