Extraction of Facial Features using an Efficient Convolution Siamese Network on Customized Dataset

Abstract

In recent years, one of the best applications of visual analysis and understanding that has drawn a lot of attention is face recognition. Due to its numerous uses in areas including safety, medical marketing, identity verification, surveillance, security etc., it has caught the interest of many researchers. In this research, a network has been proposed for face recognition. The work has been performed on a customized dataset with training and testing images. 1500 photos altogether constitute the dataset. The simulation was run using hyper parameters such as the batch-size with a value of 128. optimizer as Adam. The proposed network has outperformed with the value of accuracy as 76.48%