This document presents research on criminal recognition in CCTV surveillance videos using deep learning. It proposes a method where a user can upload faces of known criminals. When CCTV footage is recorded, the application will monitor for these faces. If a face is recognized, the CCTV camera will track the identified person through multiple cameras by alerting other cameras. The system segments video into images, acquires images, recognizes human faces, constructs motion flows between cameras to track individuals. Experimental results on a dataset show the system's ability to extract patterns from faces and cluster images of different angled faces. The system aims to identify criminals across surveillance camera networks.