FACE SKETCH RECOGNITION USING DEEP LEARNING

Abstract

Face Recognition is used to recognise or identify the faces through photos or videos. Today, Face Recognition has a very wide application involving various fields. The accuracy, speed and easy implementation behind the facial recognition system makes it even more useful in day-to-day work. Beyond just unlocking phones and laptops, the face recognition and identification is highly used among security and surveillance. Recognition of face from sketch is very important in police verification to track any person or criminals. The face is drawn by sketch artist as described by the eyewitness and then the face is recognised through the police database. There are various methods to automatically identify the sketches from a large database has been implemented but using the forensic sketches often reduce the performance. Many techniques have been used to automatically identify subjects described by eyewitnesses in sketches; however, these techniques frequently perform worse when extended galleries resembling law enforcement mug-shot galleries and real-world forensic sketches are used. Despite deep learning's success in many application areas, including traditional face recognition, not much effort has been done to apply it to face photo-sketch recognition. This is mostly because there aren't enough sketch pictures accessible to train big networks in a reliable manner. Using deep learning techniques, this thesis attempts to address these problems.