Artificial intelligence has been developing rapidly and now enables us to solve problems such as counting traffic on roads. However, conditions for vehicle detection and classification can vary, influenced by factors like the angle of view, weather conditions, and different types of vehicles. We have implemented a method that counts vehicles in footage regardless of weather conditions. Several detection models were used, including YOLOv8, SSD, and RT-DETR. All models were pre-trained on the COCO dataset and further trained on data compiled from various smaller datasets. Based on the results, we concluded that YOLOv8 and RT-DETR are the most effective models for detecting and counting vehicles in traffic.
|
