earticle

영어

This study aims to determine the optimal YOLO algorithm for object detection performance in PreScan simulation. YOLOv4, the current model used, is compared with five versions of YOLOv8 (v8n, v8s, v8m, v8l, v8x). The analysis focuses on simulation time, initial detection distance, and continuous detection distance to assess the suitability of each version for practical applications. YOLOv8n, designed for speed and efficiency, demonstrated significant improvements in reducing simulation time, while YOLOv8x showed exceptional performance in terms of detection accuracy and consistency. The various versions of YOLOv8 each present unique strengths, with YOLOv8n excelling in speed and YOLOv8x in accuracy. YOLOv8m offered a well-balanced approach, making it particularly suitable for scenarios where both speed and detection reliability are crucial. The study highlights the potential benefits of adopting YOLOv8 models in PreScan simulations, emphasizing their ability to enhance object detection performance in the tested scenario. While further testing is needed to confirm applicability across various conditions, the results suggest that each version has unique strengths that could be leveraged depending on specific requirements. Overall, YOLOv8m is identified as the optimal choice for balanced performance in the current test scenario, ensuring both effective simulation and reliable object detection capabilities.