Convolutional Neural Network for Overcrowded Public Transportation Pickup Truck Detection

Thailand has been on the World Health Organization (WHO)’s notorious deadliest road list for several years, currently ranking eighth list. Among all types of fatalities, pickup trucks converted into vehicles public transportation are found to be most problematic due their high occupancy and minimal passenger safety measures, such as belts. Passenger overloading is illegal, but it often overlooked. The country uses police checkpoints enforce traffic laws. However, there few or no highway patrols apprehend offending drivers. Therefore, in this study, we propose use existing closed-circuit television (CCTV) cameras with deep learning techniques classify overloaded transport (PTPT) help reduce accidents. As said type vehicle its characteristics unique, a new model deemed necessary. contributions study follows: First, used various state-of-the-art object detection YOLOv5 (You Only Look Once) models obtain optimum overcrowded pretrained our manually labeled dataset. Second, made custom dataset available. Upon investigation, compared latest discovered that YOLOv5L yielded optimal performance mean average precision (mAP) 95.1% an inference time 33 frames per second (FPS) graphic processing unit (GPU). We aim deploy selected control computers alert passenger-overloading violations. chosen algorithm feasible expected traffic-related fatalities.