Facial recognition and detection using Convolution Neural Networks

Pattern and facial recognition solutions are one of the biometric recognition technologies that are based on processing information of facial features. Here, we aimed to develop a classification model for face recognition of 400 images collected in the AT&T face (ORL) database. The set images were taken in the laboratory between April 1992 and April 1994 from 40 distinct people with several lighting, facial expressions, and facial characteristics. The developed model is based on the improved technique CNN and trained on 400 images of 40 persons in 10 different aspects (lighting, glasses, smiling, closed/open eyes). The accuracy achieved by the developed CNN model reached 99.04% for the majority of classes. To conclude, the proposed classification models provided important information with the highest accuracy for face recognition. These developed models and strategies could lead to the recognition of faces for several reasons (security, matching, recognition, and so forth) to detect criminals, unknown employees, or attendees.