Evaluation of deep learning models for oil tank detection utilizing SPOT imagery
Mukesh Dalal, Payal Mittal, Vedant Goyal, Rushil Updahyay, Naman Sood · 2025
The accurate and efficient detection of oil tanks from satellite imagery is crucial for monitoring and managing energy resources, as well as for various other applications such as urban planning and infrastructure management. A substantial quantity of satellite imagery has recently become accessible, utilized in both military and civilian applications. The enhanced resolution of the new spaceborne sensors facilitates the identification of specific objects of differing sizes. In recent years, deep learning-based models have demonstrated impressive performance in a wide range of computer vision tasks, including the detection and classification of objects in remote sensing imagery. Consequently, remote sensing equipment serves as optimal tools for the detection of oil tanks. Concerning this matter, this paper implemented the latest four YOLO models explicitly YOLO8x, YOLO9e, YOLO10x and YOLO11x for the detection of oil tanks using high-resolution optical imaging. The obtained results show that YOLO10x and YOLO11x models deliver more precise detections with mean average precision (mAP) of 95.90% in both cases. The qualitative and quantitative assessments divulge that these models provide comprehensive, transferable, and efficient results.