Object detection in remote sensing imagery is crucial for numerous applications such as urban planning, environmental monitoring, and disaster response. This paper investigates the effectiveness of transfer learning using YOLOv8 for object detection in remote sensing, specifically focusing on the DIOR and Ships datasets. We compare models trained from scratch with those pre-trained on the COCO dataset and subsequently fine-tuned on DIOR. Our experiments demonstrate that the pre-trained YOLOv8 model significantly outperforms the scratch-trained model across all evaluated metrics, including precision, recall, and mean Average Precision (mAP). The pre-trained model demonstrates enhanced detection capabilities and reduced misclassifications. Detailed analysis using confusion matrices highlights the model’s improved ability to distinguish between visually similar categories and detect smaller, less distinct objects. These findings underscore the advantages of leveraging transfer learning to enhance the performance of object detection models in remote sensing imagery. Our study concludes that fine-tuning pre-trained models on specific datasets leads to robust and accurate detection, offering a valuable approach for remote sensing applications.