Abstract: Aiming at the problem of high similarity between different types of remote sensing image scenes and large differences within the same type, which leads to the easy misclassification of similar scenes, this paper proposes a classification method of remote sensing scenes with a bilinear attention mechanism. The lightweight InceptionV3 was selected as the backbone network. The attention mechanism module was introduced to highlight salient feature information and ignore irrelevant background information. The transfer learning was introduced to improve the performance of the model for scene classification. The results show that the classification accuracy of our method reaches 97.50% and 96.70% on the NWPU45 and AID datasets, respectively. It is demonstrated that the proposed model can improve the performance of remote sensing images scene classification.