Congratulations to Xinfu Liu on ACM JDIQ Paper Acceptance

A Remote Sensing Image Classification Method Based on Detail Attention Sampling and Teacher-Student Network ACM Journal of Data and Information Quality （中科院4区）

With the continuous development of remote sensing technology, high-resolution remote sensing images have the characteristics of wide coverage, diverse target changes, and complex backgrounds, and the amount of their data is increasing day by day. However, the receptive field of the current convolutional neural network is relatively small, making it difficult to capture global contextual information. To address this issue, we propose a remote sensing image classification method that combines the detail attention mechanism and the teacher-student network (DATS) to effectively obtain global contextual information. Firstly, through the detail attention mechanism, the spatial relationships of the feature maps are integrated into the feature channels, thereby converting the feature maps into attention maps to generate images that maintain both structure and details. Subsequently, the teacher-student network takes the detail-preserving and structure-preserving images as inputs, and utilizes the feature refiner to enhance the fine-grained details of the images. Finally, through knowledge distillation, the fine-grained details learned by the teacher network are integrated into the main network to achieve the effective fusion of local detailed features and global structural features. Experiments on the FGSCR-42, WHU-RS19, and NWPU datasets show that the Top-1 classification accuracy of this method reaches 88.82%, 91.82%, and 87.60%, respectively.