Abstract

Small object detection in remote sensing images is essential for urban planning, environmental monitoring, disaster relief operations, and defence. Detecting small objects is still challenging because of sparse pixel representation, class imbalance, and complex background. The conventional object detection models, such as Faster R CNN and RetinaNet, have difficulty preserving spatial information, and the detection accuracy is lower. In this paper, we apply YOLOv8, a cutting-edge deep learning model, to improve detection of small objects. The model is trained and tested on the DIOR data set, which consists of diverse aerial images with horizontal bounding boxes. Preprocessing operations involve parsing XML annotations, translating bounding box coordinates to YOLO format, and performing data augmentation to enhance generalization. The performance of the model is measured using mean Average Precision (mAP), Precision, Recall, and Accuracy. Experimental results show that YOLOv8 obtains a mAP@50 of 73.4% on training and 71.0% on testing, and a mAP@50–95 of 50.3% and 48.5%, respectively. The model also obtains high Precision and Accuracy, better than the previous versions of YOLO and conventional detectors. In comparison with transformer-based models such as DETR, YOLOv8 provides the best speed-accuracy trade-off for real applications. This work forms an effective basis for detecting small objects, allowing scalable and automated surveillance systems for high-resolution satellite images in remote sensing applications.