VGG FaceNet Based Sketch to Face Recognition with Morphable Model

Abstract

Sketch to face recognition automation can play important role in forensic operations. The forensic departments can generate sketches with the help of drawing artists. The resulting sketch images may have difference compared to actual faces in terms of facial parts and expressions. The convolutional neural network (CNN) based method proposed in this paper shows augmentation based sketch and facial expression dataset generation by modifying the public dataset. The generated dataset is thus used to train the VGGFaceNet CNN model and performance is evaluated. The performance of VGGFaceNet model is tested with reference to parameters like accuracy, specificity and sensitivity. The proposed system indicates accuracy of 88% over to other conventional methods such as Local Binary Pattern, Support Vector Machine.