Research on Enhancing High-Resolution Remote Sensing Object Detection Performance Based on C2FEMBC and GFPN

The growing demand for high-resolution imagery across various application domains has led to a need for advanced object detection models that can handle complex scenes with improved accuracy and efficiency. This paper presents novel enhancements to the YOLOv8 model by incorporating Cross-Stage Feature Enhancement with Modified Bottleneck Convolution (C2FEMBC) and Generalized Feature Pyramid Network (GFPN). These enhancements are designed to optimize feature extraction and multi-scale representation, addressing key challenges in detecting small objects and complex patterns. The modified model is evaluated using a high-resolution dataset containing diverse object categories, demonstrating a significant improvement in detection performance. Notably, the proposed modifications result in a 59.4% increase in the mAP50:95 metric compared to the baseline YOLOv8 model. The research highlights the importance of leveraging advanced architectural modifications to enhance detection accuracy in high-resolution imagery, offering a promising approach to address the limitations of current object detection techniques. These advancements provide valuable insights into improving object detection models for real-world applications, where precision and robustness are critical.