Real-time face identification from video surveillance cameras

In this research work, we present the implementation of a web environment and processing core based on convolutional neural networks, to recognize faces in real time. The environment can work with several video surveillance cameras at the same time. We conclude that ten images of a person are required for recognition to be highly reliable. The environment was tested with 32 volunteers, in 4 different environments: front face, smiling, wearing a cap, and glasses. We obtained 94%, 91%, and 90% in metrics true-positive-rate, precision, and accuracy, respectively.