DEVELOPING A VEHICLE IDENTIFICATION SYSTEM IN VIDEO USING YOLOV10

Abstract

Object recognition, a popular topic in the field of machine learning. In this thesis, we are conducting experiments with the machine learning model YOLOv10 to investigate its capabilities in vehicle identification, specifically focusing on realtime video analysis, customizable identification regions in video, and vehicle counting. Consequently, we assess the performance of the model in recognizing Vietnamese means of transport by employing a custom dataset comprising 3000 images categorized into 8 classes. The results indicate that the model trained with YOLOv10 performs well under standard conditions and new features, achieving a Precision of 0.951, Recall of 0.933, mAP50 of 0.976, and mAP50-95 of 0.888 respectively.