Integration of Generative Adversarial Networks (GAN) and AI Drawing in Criminal Sketches—Applied in Crime Scene Investigation by Law Enforcement

Conventional criminal sketch techniques often rely on the identification memory of eyewitnesses, introducing subjective factors. However, with the evolution of the AI era, AI drawing has emerged as a supportive tool. This research aims to explore the use of Generative Adversarial Networks (GAN) to assist criminal sketch artists in enhancing the likelihood of identifying suspects, applied in crime scene investigations by law enforcement. In this paper, we employ a GAN model to develop a system that automatically generates sketch images using Precise Prompting (PROMPT). This system can produce a realistic criminal sketch based on the criminal features provided by the police. Experimental results demonstrate that our proposed method achieves a similarity score of 0.63 in generating criminal sketches, a value already perceptible to the naked eye. Furthermore, compared to traditional sketching techniques, our approach exhibits higher efficiency and reliability. Therefore, the outcomes of this study hold promise for application in police investigations, improving the efficiency and accuracy of crime resolution.