Many developments have taken place in the field of face-recognition and liveness analysis to improvise various device securities and attendance verification systems. Many approaches have incorporated 3D scans of the face to predict the liveness of the person in front of it. Our method of analysis tries to account for this problem without using advanced 3D imaging techniques or hardware. It consists of two parts, the former helps in face verification and the latter to check the liveness of the face in front of it.
In the first part, we have used a model based on Google’s FaceNet (Inception) Model which learns a mapping from face images to compact Euclidean space distances, which directly correspond to the measure of similarity of the images. Once the space has been produced, face verification can be easily implemented using standard techniques with embeddings as feature vectors.
For the second part, we have employed a cascaded multi-task framework that extracts certain features from the facial image which are then used to check for liveness by tracking their relative displacements. In this case, Multi-Task Cascaded Convolutional Network (MTCNN) has been used. These extracted features were used to check the liveness of the person’s face by asking them to perform some tasks in a random order like moving their head towards right or smile etc.
