The paper introduces GALDET, a novel method for remote sensing object detection that addresses the challenges of detecting small targets in high-resolution images. The method integrates a Global and Local Attention Mechanism (GAL) to effectively capture both fine-grained details and global contextual information. The GAL mechanism is implemented through a feature fusion module (CGAL), which combines local and global attention to enhance the model's understanding of complex images. Additionally, the method features a multi-decoupled prediction head structure to handle multi-scale features and a novel Ziou loss function to improve the accuracy of small target localization. Experimental results on the VisDrone2019 and DOTA datasets demonstrate that GALDET significantly outperforms traditional methods, achieving a 6.3% to 6.5% improvement in Average Precision (AP) for small target detection. The paper also discusses the limitations and future directions for the method, emphasizing the need for further refinement in detection accuracy and real-time processing speed.The paper introduces GALDET, a novel method for remote sensing object detection that addresses the challenges of detecting small targets in high-resolution images. The method integrates a Global and Local Attention Mechanism (GAL) to effectively capture both fine-grained details and global contextual information. The GAL mechanism is implemented through a feature fusion module (CGAL), which combines local and global attention to enhance the model's understanding of complex images. Additionally, the method features a multi-decoupled prediction head structure to handle multi-scale features and a novel Ziou loss function to improve the accuracy of small target localization. Experimental results on the VisDrone2019 and DOTA datasets demonstrate that GALDET significantly outperforms traditional methods, achieving a 6.3% to 6.5% improvement in Average Precision (AP) for small target detection. The paper also discusses the limitations and future directions for the method, emphasizing the need for further refinement in detection accuracy and real-time processing speed.