Distinguish Traffic Condition Based on YOLOv10 Model and Region of Interest (ROI)

Traffic condition estimation and distinguishing from real‐time video cameras are essential research in traffic information systems (ITSs), providing accurate and live traffic information to commuters and traffic management, as video‐based traffic data are available ubiquitously. However, many challenges need to be resolved, including the complexity of video data with multiple vehicle types in chaos traffic environments, such as those in developing countries like Vietnam, the lack of specific training datasets, and the appropriate selection or modification of pretrained deep learning (DL) models for video data processing. This paper proposes a novel traffic congestion prediction method based on real‐time traffic video analysis utilizing appropriate DL models. The proposed approach using YOLOv10, YOLOv8, and Faster R‐CNN to detect and classify vehicles and region of interest (ROI) to calculate the occupied their area, which resulted in a prototype system for real‐world applications consisting of three main stages: (i) traffic video data in a chaos traffic environment, specifically in Hanoi, Vietnam, are collected, preprocessed, and annotated for traffic conditions; (ii) various pretrained DL models for video data analysis specified to traffic condition estimations are studied to apply to the above traffic video data; and (iii) thorough evaluations using the implemented prototype with real‐time video traffic data to confirm the effectiveness and the efficiency of the proposed method have been analyzed. The results indicate that the proposed method achieves up to 94% accuracy in vehicle detection and processes at a speed of 27 frames per second. The implemented prototype also provides a visual presentation of traffic density and makes reliable congestion predictions to commuters and management. The proposed approach not only supports traffic operation and management in regulating traffic flows but also paves the way for applying technology to address complex urban traffic challenges, especially in developing countries.