Face Recognition with Convolutional Neural Network and Transfer Learning

Face recognition finds its major applications in biometrical security, health care and marketing. Among many biometrics like finger print, iris, voice, hand geometry, signature etc., face recognition has attained the most popular research orientation because of high recognition rate, uniqueness and large number of features. Some of the challenges in face recognition task include the differences in illumination, poses, expressions and background. Now-a-days deep learning methods are widely employed and they ensure promising results for image recognition and classification problems. Convolutional Neural Networks automatically learn features at multiple levels of abstraction through back propagation with convolution layers, pooling layers and fully connected layers. In this work, an automated face recognition method using Convolutional Neural Network (CNN) with transfer learning approach is proposed. The CNN with weights learned from pre-trained model VGG-16 on huge ImageNet database is used to train the images from the face database. The extracted features are fed as input to the Fully connected layer and softmax activation for classification. Two publicly available databases of face images - Yale and AT&T are used to test the performance of the proposed method. Experimental results verify that the method gives better recognition results compared to other methods.