YOLOv8: A Comprehensive Review of Architecture, Advancements, and Applications in Real-Time Object Detection

Abstract - Recent advancements in deep learning have significantly improved object detection systems, with the YOLO (You Only Look Once) family emerging as a dominant paradigm for real-time detection. YOLOv8, developed by Ultralytics, introduces architectural refinements such as anchor-free detection, decoupled heads, and multi-task capabilities. This paper presents a comprehensive review of YOLOv8, including its architecture, innovations, performance benchmarks, and applications. Furthermore, the study compares YOLOv8 with prior YOLO versions and other detection frameworks while highlighting limitations and future research directions.

Key Words: object detection, computer vision, YOLOv8, YOLO, deep learning, decoupled head, convolutional neural networks (CNNs), feature pyramid networks (FPN), image recognition, edge computing