Application du Deep Learning en vidéo surveillance

Abstract

The objective of this end-of-study project is to create a system capable of counting crowds and and verify the faces of suspect people present in real-time video surveillance using deep learning. This project is divided into two parts, in the first part the yolov3 model is used to detect and count people, and the second part, it uses the FaceNet model in order to detect the faces of suspect people and we also train it on images, assuming they are images of suspect people, we obtained satisfactory results and finally, we combine these two parts in the graphic interface.