Enhancing Convolutional Neural Networks for Face Recognition with Occlusion Maps and Batch Triplet Loss

Despite the recent success of convolutional neural networks for computer applications, unconstrained face recognition remains a challenge. In work, we make two contributions to the field. Firstly, we consider the of face recognition with partial occlusions and show how current might suffer significant performance degradation when dealing with kind of face images. We propose a simple method to find out which parts of human face are more important to achieve a high recognition rate, and use information during training to force a convolutional neural network to discriminative features from all the face regions more equally, including that typical approaches tend to pay less attention to. We test the of the proposed method when dealing with real-life occlusions using AR face database. Secondly, we propose a novel loss function called batch loss that improves the performance of the triplet loss by adding an term to the loss function to cause minimisation of the standard deviation both positive and negative scores. We show consistent improvement in the Faces in the Wild (LFW) benchmark by applying both proposed adjustments the convolutional neural network training.