YOLO 시리즈(V1에서 V11까지)와 응용 애플리케이션 분석 비교

Abstract

Deep Learning has emerged as multi-domain innovation aimed the popularity of various Machine Learning techniques, and YOLO makes it possible for widely using in deep learning for object detection tasks. In this paper, we present a analysis of YOLO series (original scheme to version 11), undertaking a comprehensive analysis of YOLO’s performance and synthesizes existing researches. We start by describing the architecture and contribution of YOLO families, discussing the major changes in network architecture, and training tricks for each model. Then, se summarize the necessary lessons form YOLO’s development and provide a perspective on its highlighting research directions to enhance object detection.