The Application and Performance Comparison of Different Versions of YOLO Image Recognition Systems †

Abstract

With advancements in computational power and artificial intelligence (AI), image recognition has become more efficient and widely used. You Only Look Once (YOLO) stands out for its fast and accurate object detection, making it popular among researchers. The technology enhances daily life, from smartphone facial recognition improving security to thermal imaging aiding public health during the pandemic. However, identifying the right image recognition system for varied image types remains challenging. By evaluating the performance of YOLO’s versions (e.g., YOLOv3 and YOLOv4) regarding structure, speed, accuracy, and adaptability, we identified appropriate algorithms for specific tasks, recommending optimal image recognition techniques for different applications.