Suspect Face Generation and Recognition Based on DCGAN

Traditional face generation requires the sketch artist to create image of the suspect based on the information. It consumes time and the clarity of the image is also not appropriate. Also, the existing traditional face recognition methods in suspect detection often struggle with limitations in image quality or disguise. To overcome the limitation in suspect face detection, this paper explores the application of Generative Adversarial Networks (GANs). This research work proposes an improved GAN architecture, training process, and evaluation methods for suspect's face detection. To improve the accuracy of face recognition, the proposed approach is implemented in an application that allows users to generate face and recognize the correct suspect using deep learning technique without the need of forensic artists. The proposed application generates realistic facial images of suspects based on partial information or descriptions such as variations in appearance, such as hairstyles, facial hair, or minor occlusions. This enables matching suspects even with low-quality footage or attempts to alter their appearance.