Facial Point Detection via Deep Neural Networks

Face recognition is one of the most significant branches in computer vision research. the most fundamental but by far the most important task is facial keypoints detection, that is, to find out the locations of specific keypoints on face images In this thesis,, we are given a list of 96x96-pixel 8-bit graylevel images. The task is to predict the positions of 15 keypoints on grayscale face images. Each predicted keypoint is specified by an (x, y) real-valued pair. In experiments, we trained Deep Convolutional Network (CNN) and varied the depth and size of an architecture. The experimental results show that the network-4 which has 7 layers with the provides better results than other's model for the dataset.