Object Detection in Remote Sensing Images Using YOLOv8

Detecting objects in remote sensing images (RSIs) is essential for various applications, from urban planning to disaster response. However, current deep learning models often struggle due to their reliance on fixed anchor boxes and the lack of object-contextual clues within the images. A specialized object detection system has been developed using YOLOv8 technology for analyzing remote sensing images. The framework is designed to accurately identify objects within complex satellite imagery. By leveraging the DIOR dataset and meticulously converting conventional annotations into the YOLO format, the model undergoes rigorous training, culminating in substantial accuracy gains. Diverging from conventional methodologies, the framework places paramount importance on the integration of scene-contextual features and the implementation of cutting-edge data augmentation strategies. This deliberate emphasis on contextual comprehension and exposure to diverse training data is anticipated to significantly amplify detection precision and bolster the model's adaptability across a wide spectrum of remote sensing image scenarios. Performance and efficiency metrics of proposed model were significantly increased compared to YOLO CSL model and current state-of-the-art models.