YOLO-Based Object Detection: Evolution, Real-Time Performance, and Applications in Intelligent Vision Systems

The YOLO (You Only Look Once) family of object detection algorithms has transformed the field of computer
vision by enabling real-time, high-accuracy detection in diverse application scenarios. This review presents a
comprehensive review of the architectural evolution of YOLO from the foundational YOLOv1 to the recent
YOLOv8 emphasizing innovations such as anchor-free detection, multi-scale fusion, dynamic heads, and
transformer-aware modules. Comparative evaluations against classical detectors like Faster R-CNN and SSD
highlight YOLO’s unparalleled balance between inference speed and detection precision, particularly in
resource-constrained and embedded environments. The paper further explores YOLO’s practical deployments
in autonomous driving, smart surveillance, medical diagnostics, industrial automation, and agriculture.
Benchmarking comparison across datasets such as COCO, KITTI, and PASCAL VOC are discussed alongside
evaluation metrics like mean Average Precision (mAP), Intersection over Union (IoU), and inference latency.
Key challenges including small object detection, domain adaptation, and explainability are examined, along
with future directions involving edge-optimized deployment, multimodal integration, and ethical AI design. By
consolidating architectural, empirical, and domain-specific perspectives, this review aims to serve as a
foundational resource for researchers, engineers, and practitioners seeking to harness the power of YOLO in
real-world intelligent vision systems.