Facial Description Based Suspect Detection
Kiran S Muragodnavar, K J Jagruthi, Neha V Revankar, Konka Kishan, Surabhi Narayan · 2024
Suspect identification poses a significant challenge for investigators in day-to-day crimes due to limitations in available data from CCTV footage and eyewitness descriptions. The obtained images may not always be clear, and not all areas are covered by surveillance cameras. Sketching faces based on eyewitness descriptions is also challenging due to the scarcity of skilled artists and the time-consuming nature of the process. This research introduces the Facial Description-Based Suspect Detection (FDBSD) system for identifying suspects using facial descriptions provided by eyewitnesses and CCTV images. The proposed model is tested on the IDOC Mugshot dataset, containing details and facial images of 70,008 criminals. Google Teachable Machine and Keras models were employed to identify various facial features. A graph representation is built based on similar facial attributes and facial similarity. The accuracy of the proposed model is 91.7 %.