Comparative Analysis of Face Recognition Models for Criminal Detection

Facial recognition technology has gained significant attention in recent years for its potential in improving criminal detection and enhancing security systems. This paper compares and contrasts different facial recognition technologies which are being implemented in video datasets for criminal detection. Law enforcement organizations are vigilantly monitoring facial recognition technologies as an approach to finding and apprehending criminals. The objective is to evaluate and compare how different deep learning techniques perform in real criminal detecting circumstances. A comprehensive video dataset that includes footage from law enforcement archives and surveillance camera footage is curated. To ensure a representative sample for the study, the dataset consists of a group of people with an assortment of physical attributes and notoriety. Each facial recognition technology is used and refined through transfer learning on the curated video dataset. To examine the performance of the systems, accuracy, precision, and error rate are employed as performance assessment variables. The comparative analysis highlights both the advantages and disadvantages of each approach. VGGFace, FaceNet, OpenFace, and DeepFace models are being assessed where the VGGFace model depicted the highest accuracy. The findings offer valuable insights into the performance and applications of various facial recognition systems. The findings expand facial recognition technology and demonstrate the value of law enforcement tools in strengthening public safety.