Creation of Face for Aiding in Forensic Investigation based on Textual Description

In this work, we have developed a solution that will assist the police department in the criminal investigation process during drawing sketches of a suspect based on the textual description from a witness. We propose a deep learning implementation based on Generative Adversarial Network that consists of generator which generates an image based on text description and discriminator which validates the image generated by generator with the data set. We trained a subset of 400 images collected from the Labeled Face In Wild dataset. The description of images was taken from Face2Text data provided by Albert Gatt. The images generated resemble certain features mentioned in the textual description which can directly aid the cops or can act as an intermediary for sketch artists to work upon.