Face Alignment And Face Verifiation Based On Deep Learning

In recent years,with the development of internet and algorithms in computer vision,the image processing technologies based on deep learning gain their popularity.As the main part of computer vision,face alignment and face verification also benefit from that.In this paper,we propose an adaptive cascade framework for face alignment,termed Adaptive Cascade Deep Convolutional Neural Network.In this system,Gaussian distribution is used to bridge the current network input with the previous network output.A new system for face verification is also presented.Our contributions are summarized as follows:1.This dissertation summarizes the Supervised Descend Method and points that the cascade system is effective and efficient for face alignment.2.A new network structure for facial point regression is introduced.And an adaptive algorithm for training samples selection is presented based on Gaussian distribution,which makes the cascade more effective.3.Two face alignment systems based on the adaptive algorithm are implemented.One is utilized to locate five facial points.The other is employed to detect sixty eight facial points.4.An approach to extract face features based on deep convolutional neural networks and face alignment is presented.A new system for face verification is also introduced.97.2%accuracy is achieved on the public available LFW benchmark.