Transfer Learning with Deep Convolutional Neural Networks in Forensic Face Sketch Recognition

Abstract

Over the last few years, there has been tremendous progress in the field of artificial intelligence, and crime rates have risen at an alarming rate. As a result, the demand for systems that can identify, detect, and recognise suspects from sketches of their faces has grown. Some automatic methods for identifying suspects from sketches have been developed before, but their efficiencies are insufficient for both real-world and software-generated sketches. Face sketch to photo-realistic picture conversion using Deep Convolutional Neural Network model and face identification using pre-trained VGG-Face model by applying transfer learning are shown in this paper. Convolutional neural networks are used because of their vast data processing power and hierarchical feature representations. The algorithm outperformed existing methods in face sketch recognition, recognising the identity of people from sketches with an average accuracy of 0.98. This method could be extremely beneficial in forensic investigation for identifying suspects.