Crime Investigation using DCGAN by Forensic Sketch-to-Face Transformation (STF)- A Review
S. Nikkath Bushra, K. Uma Maheswari · 2021
This paper outlines about the most advanced technique of Artificial Intelligence for digitally ascertaining a criminal through facial recognition system by converting forensic sketch into a real photo using Deep Convolutional Generative Adversarial Network (DCGAN). Suppose a crime act is reported to a cop as an eyewitness by an individual by remembering certain set of facial features of an illicit and trying to imitate it in the form of hand drawn sketch based on the given information. The forensic sketch is pictured by an expert based on the verbal explanation given by a person after commitment of crime by a perpetrator. The rough sketch is given as an input to train the neural network and after several epochs the network quickly learns and generates a scrupulous realistic facial image of a suspect from the forensic sketch. This is very much helpful in crime investigations to obtain several such real photographs of a suspect from forensic sketches easily with precise details within a short period of time. This is an amazing practice that can produce real facial images of high resolution color photos from a low quality sketches which is incomplete or partial in nature with different pose variations like color, tone etc. It is very much useful in domains like forensics, law enforcement, Facial recognition system and security and authentication systems.