Key points are not available for this paper at this time.
Converting pencil sketches to real image is required in forensic analysis for identifying a person. In a reported crime, the testimony explains the physical features of the accused by recalling them from memory. As the human face is so diverse and variable, it's tough to predict the person from a forensic sketch expert's low-quality sketch. In this paper we propose a deeplearning model using DCGAN to convert the sketch to real image. The prepared sketch is used as input for training, and after a set number of epochs, our deep learning model learns and produces a realistic high-resolution image of the suspect. The face recognition system is a powerful tool which can be used for a wide range of purposes, from security to entertainment. Our proposed method's final average SSIM is 0.587 where as in the existing methods the average SSIM is 0.554.
Devakumar et al. (Sun,) studied this question.