Case Instance Segmentation of Small Farmland Based on Mask R-Cnn of Feature Pyramid Network with Double Attention Mechanism in High Resolution Satellite Images

•Two new attention modules are designed, spatial attention and channel attention.•The designed two attention modules can be easily plugged into the existing pyramid feature network.•Propose a new feature pyramid network structure for farmland segmentation.•Introduce the feature pyramid network structure based on the double attention mechanism into the Mask R-CNN framework.•The superiority of its feature extraction ability is proved by experiments.