Sketch-Based Face Recognition

Automatic facial recognition has attracted huge attention in recent years owing to rising demands in applications like law enforcement, video surveillance, banking, and security systems. People have a false opinion that only images reflect a person's face rather than sketches. In a complex environment like poor illumination, varying pose, and expressions, sketch-based recognition outperforms image-based recognition. This paper proposes a sketch-based face recognition (SBFR) method using a modified Mask RCNN framework. The proposed model includes recognition of a person even if there is a change in illumination, age, and facial expression. The method is validated with our local dataset “POPSKETCH,” and it comprises hand-drawn sketches of 28 individuals and converted sketches from original images of 34 individuals. The proposed Mask RCNN method achieves a better recognition rate of 90.67 which is higher than face recognition using the AlexNet network, YoloV5, and Mask RCNN network model.