Similarity calculation for face verification with convolutional neural network
Nishiki Katayama, Satoshi Yamane · 2017
Convolutional neural network has made major progress in classification problems of general object recognition. Classification of facial images is one of them. However, expression of the networks for classification depends on datasets and the network model, which is vulnerable to changes in the adaptation range. We propose the network that has the two convolutional parts of pre-trained CNN by transfer learning and fully-connected layers for calculating similarity. By learning the neural network that has the two convolution part of the convolutional neural network that learned the classification problem, it is possible to express a function that calculates the similarity of people by two facial images. This network was trained by back propagation method using mean square error as loss function. Consequently, the value of loss function decreased on the LFW dataset. It is evident from reduction in loss that neural network including convolutional parts can calculate the similarity of two faces. The output similarity can also be used as reliability.