Improving the Face Recognition Performance Using Gabor and VGGFace2 Features Concatenation

This paper presents a new method for face recognition that is based on concatenation of Gabor and VGGFace2 features. The proposed method is composed of three stages: a feature extraction, concatenation and a recognition stage. The feature extraction stage is implemented using the pre-trained model VGGFace2 and PCA is used to reduce the Gabor features, while the recognition stage is implemented using full connected neural networks. The proposed method was evaluated on the IJB-A dataset and Labeled Faces in the Wild (LFW) dataset. The results show that it is able to achieve recognition accuracy more than 99%.