Gabor-Deep CNN based Masked Face Recognition for Fraud Prevention

With the new overall COVID-19 pandemic, utilizing face veils have become a significant piece of our lives. To evade the spread of contamination Individuals are urged to cover their faces using a mask when in open territory. The utilization of these face veils makes the orthodox facial recognition system inadequate for authentication in school/office participation, face access control, facial participation, facial security checks at airport and train stations, etc. The recent progressed deep learning-based face acknowledgment approaches rely upon countless training samples and significant features of the entire face. This paper addresses the issue of recognition of face and veil using Gabor wavelet and deep transfer learning. Here the Gabor wavelet features are extracted from the non-masked region of the face and integrated with deep learned CNN features to generate a more robust feature vector for optimizing recognition. The proposed method is experimented on 4 benchmark datasets and a manually prepared dates and achieved averaged 97% recognition accuracy.