YOLOV5, YOLOV7, and YOLOV8: Architectures Benchmark for 5G Pipeline Inspection Drones

The oil and gas industry heavily relies on robust transportation infrastructure, such as pipelines, necessitating regular inspection and maintenance to ensure operational integrity utilizing intelligent aerial robots. This study aims to determine the most suitable YOLO architecture for a 5G pipeline inspection drone using aerial images by comparing three state-of-the-art real-time object detection models: YOLOv5s, YOLOv7-tiny, and YOLOv8s. The results indicate that YOLOv8s outperforms other models across various metrics, demonstrating its superiority in pipeline detection tasks, while YOLOv5s exhibits faster training time, highlighting its efficiency. These findings contribute to the advancement of pipeline detection methodologies and facilitate informed decision-making in selecting optimal models for real-world deployment scenarios.