A CNN Transit Model for Object Detection using YOLOv5

The rapid urbanization and burgeoning populations in cities worldwide have led to an increased demand for efficient and sustainable urban transit systems. In India, the pace of growth in the two-wheeler fleet is 30 times greater than the rate of growth in the population. The chance of mortality is 3.5 times higher for riders who don't wear helmets than for those who do. In view of this, developed an AI-based model for object detection to guarantee public safety and enable real-time monitoring of traffic law violations. The system uses a camera to capture images of bikes carrying three or more people who are not wearing helmets as well as to recognize the license plates of bikes. A CNN Transit model for traffic networks have been proposed. The model utilizes the Yolov5 backbone layer, which is fine-tuned by increasing hyperparameters. This results in the accuracy of 86 % and a significant reduction in training time, which enhances the model's performance. By addition of PANet and FPN to the Yolov5 Neck layer has improved the detection of far and tiny objects in traffic and roadways. This study emphasizes the potential of AI-ML technology to address urban transportation difficulties and foster a culture of safety by detecting objects.