The Integration of AI enhanced face detection technologies represent a significant advancement in law enforcement technology, public safety, and crime prevention. In this work, we propose a novel system that uses cutting-edge AI models, including FaceNet, ResNet, MTCNN (Multi-Task Cascaded Convolutional Networks), and CycleGAN (Cycle-Consistent Generative Adversarial Network) to deliver highly efficient and accurate criminal identification. It is implemented as a web application with the help of Python Django framework that makes the system highly adaptable and efficient in a variety of law enforcement applications. It makes use of MTCNN and Res-Net for reliable face identification, FaceNet for accurate face embedding, and CycleGAN for accurate sketch generation. A key enhancement is the use of ArcFace loss in FaceNet to increase embedding discrimination, and CycleGAN for improved sketch-to-photo translation, expanding its utility for identifying suspects from low-quality images and sketches. The suggested method can detect offenders even in real time situations, when people are wearing masks or when photographs are blurred. Performance evaluations show that the combination of MTCNN, ResNet, and FaceNet achieves superior accuracy, F1 score, precision, and recall compared to conventional models, with validation accuracy stabilizing around 0.995. The proposed methodology provides law enforcement agencies with a powerful, efficient, and reliable tool for enhancing public safety and supporting crime prevention efforts.