A Novel Approach for Detecting Facial Key Points Using Convolution Neural Networks

Abstract

AbstractThe task of face recognition is having many real-time applications in which the process of facial keypoint detection is considered to be an intermediate and crucial step. The amount of keypoints that are using for face recognition decides the computational requirements of the algorithm. In this paper, an effort has been made to detect the useful 15 facial key points using convolutional neural networks and compared with the state-of-the-art system with 30 facial key points. We made an effort to identify the 15 facial key points (6 points from eye +4 points from eyebrows +4 points from lips +1 point from the nose) by using the proper hyperparameters for convolutional neural network. It is found that the performance of the proposed system is quite similar when compared to the system with 30 facial key points.KeywordsBatch sizeConvolutional neural networksDropout layerFacial keypoint detectionFace recognition