Crime Scene Object Detection from Surveillance Video by Using Tiny YOLO Algorithm

In recent years, law enforcement agencies and security professionals have rewarded much attention to using artificial intelligence for criminal detection based on video surveillance. The ability of deep learning (DL) models to automatically detect and follow prospective offenders saves time and money for law enforcement organization signs by allowing them to understand complicated patterns from data. This helps them conduct in-depth probes and direct their search efforts more precisely. Bladed crimes, such as swords, daggers, knives, and bayonets, and portable firearms, such as pistols, Hand gun or carbines, rifles, Kinfe and machine guns, are often found at crime scenes. In this research, a deep learning based surveillance system is proposed that is capable of identifying the presence of traced objects, such as handguns and knives, and potentially warning authorities of impending danger. Compared to DL-based object identification algorithms like the Enhanced single shot detector (ESSD) ImageNet and FRCNN (Faster Region-based convolutional neural networks), Tiny YOLO has the best real-time detection mean average precision and inference speed. Thus, proposed solution will incorporate YOLO.