YOLO Algorithm Accuracy Analysis in Detecting Amount of Vehicles at the Intersection

Abstract

The goal of this research is to find out YOLO algorithm's effectiveness on detecting the number of vehicle on road. Our activity in this research is conduct training using a dataset that we created ourselves and do traffic recording simulation in a lot of scenario using YOLO original datasets and our own datasets. The result of this research is YOLO algorithm successfully detects vehicles as much as 65.3% of the total vehicles passing on the highway and gives the wrong label as much as 20.7% of the total label given if using YOLO original dataset. YOLO algorithm successfully detects vehicles as much as 9,3% of the total vehicles passing on the highway and gives the wrong label as much as 7,4% of the total label given if using our own dataset.