Unveiling a Cutting-Edge Dataset for Oil Tank Detection: YOLO Models Put to the Test

This paper presents a comprehensive evaluation of YOLOv7 and YOLOv8 (You Only Look Once) models for oil tank detection, utilizing a novel dataset comprising 12,948 images and 171,809 annotations. The dataset, which is the largest and most diverse in the domain, combines publicly available resources with high-resolution captures from Google Earth, significantly enhancing the robustness and generalizability of the models. The study rigorously assesses the performance of various YOLOv7 and YOLOv8 variants in challenging real-world scenarios. Notably, YOLOv8x achieves a high precision of 0.9153, while YOLOv7 standard delivers a precision of 0.9241 and an <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F$</tex>1-score of 0.8715, demonstrating superior detection capabilities across different metrics. The deployment of these models on the NVIDIA Jetson Xavier NX, achieving up to 14.09 frames per second (FPS) with YOLOv8n, marks the first instance of edge device deployment for oil tank detection, demonstrating the models' capability to deliver real-time monitoring solutions. This work sets a new benchmark for oil tank detection, offering valuable insights for future research and practical applications.