Training Siamese Neural Network Using Triplet Loss with Augmented Facial Alignment Dataset

Abstract

In recent years, deep learning methods, especially CNN, have been gaining huge progress in the development of technologies and humanity. Despite this progress, face recognition challenges are still hindering it. In this paper, we investigate the improvement in the performance of face recognition models by applying a Siamese neural network with triplet loss function and train with an augmented facial dataset. Furthermore, this dataset is collected, cropped, aligned, and augmented with various adjustments in which fill the facial recognition challenges requirements. Moreover, we compare the proposed model with the two best public models using two proposed algorithms. Experimental results display good improvement, and we discuss the possible usage as in checking attendance or biotechnique.