AIGuard: Criminal Tracking in CCTV Footage Using MTCNN and ResNet

Abstract

The identification of criminals from closed circuit television (CCTV) feeds manually is a tedious task for police personnel. It is not possible for them to continuously monitor each and every frame of footage efficiently which causes a low probability of tracking a specific criminal from it. Today, in the era of artificial intelligence, there is a prime need for an automatic visual surveillance system to solve the above issue. So, we have proposed a framework for automatically tracking a specific criminal in CCTV footage. The proposed framework uses cutting-edge algorithm such as MTCNN for detecting the face in the frames of footage and uses the ResNet model having 29 convolution layers for recognizing the specific criminals from the detected faces. The experimental results show that the proposed system is highly efficient in recognizing criminals in CCTV feeds. The accuracy of the proposed framework in detecting and recognizing the faces from the front is 85.9% and while faces having not proper front is 56.1 %.