线性注意力与跨尺度融合的遥感语义分割网络

张景鸿,邵玉斌,唐菁敏,杜庆治,龙华,张赜涛

Remote sensing semantic segmentation network with linear attention and cross-scale fusion

Jinghong ZHANG,Yubin SHAO,Jingmin TANG,Qingzhi DU,Hua LONG,Zetao ZHANG

表 1 Potsdam测试集上不同方法的定量性能对比

Tab.1 Quantitative performance comparison of different methods on Potsdam test set %

方法 骨干网络 F1 平均F1 OA mIoU 不透水地表 建筑 低矮植被 高植被 车辆 EaNet ResNet18 92.0 95.7 84.3 85.7 95.1 90.6 88.7 83.4 MAResU-Net ResNet18 91.4 95.6 85.8 86.6 93.3 90.5 89.0 83.9 SwiftNet ResNet18 91.8 95.9 85.7 86.8 94.5 91.0 89.3 83.8 FANet ResNet18 92.0 96.1 86.0 87.8 94.5 91.3 89.8 84.2 ABCNet ResNet18 93.5 96.9 87.9 89.1 95.8 92.7 91.3 86.5 Segmenter ViT-Tiny 91.5 95.3 85.4 85.0 88.5 89.2 88.7 80.7 BANet ResT-Lite 93.3 96.7 87.4 89.1 96.0 92.5 91.0 86.3 SwinUperNet Swin-Tiny 93.2 96.4 87.6 88.6 95.4 92.2 90.9 85.8 UNetFormer ResNet18 93.6 97.2 87.7 88.9 96.5 92.8 91.3 86.8 DSCT Swin-Tiny 87.8 93.0 78.0 78.6 76.2 91.1 90.7 85.7 MobileViT V3 MobileViTv3 88.9 93.7 79.0 80.2 92.1 92.6 91.1 86.8 ORLA-OCM UNetFormer(本文方法) ResNet18 94.1 97.5 87.8 89.1 96.3 93.0 91.7 87.1