DISTANCE AND ACCURACY IN OBJECT DETECTION BASED ON YOLOV8 COMPUTER VISION ALGORITHM

Abstract

Artificial intelligence is on the rise and has undergone massive growth in the industry, especially in computer vision. The emergence of computer vision from autonomous cars, robotics, surveillance, and many more has led to challenge Artificial intelligence's confidence accuracy in detecting an object. Many artificial intelligence algorithms are used by the industry, one of them is You Only Look Once version 8 (YOLOv8). YOLOv8 is a deep learning model for object detection. YOLOv8, which is developed by Joseph Redmon and Ali Farhadi is a powerful method to detect an object in real time because YOLOv8 has the capability of processing high-resolution images at high speeds. The research discusses about accuracy YOLOv8 in detecting object with a certain distance from far to very close distance. The dataset is used to train the model of YOLOv8. The dataset is collected by taking photos of an object with constant lighting but different distances. This research aims to obtain the most effective distance that the YOLOv8 computer vision algorithm model could detect. The hipothessis is there is connection between distance and detection accuracy of YOLOv8. If the distance increases, the detection accuracy decreases. However, if the object is close the detection accuracy increases. So based on the results, a conclussion could be concluded that a YOLOv8 model would have the highest accuracy at a certain distance.