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 2 Vaihingen测试集上不同方法的性能比较

Tab.2 Performance comparison of different methods on Vaihingen 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 FANet ResNet18 90.7 93.8 82.6 88.6 71.6 85.4 88.9 75.6 EaNet ResNet18 91.7 94.5 83.1 89.2 80.0 87.7 89.7 78.7 MAResU-Net ResNet18 92.0 95.0 83.7 89.3 78.3 87.7 90.1 78.6 SwiftNet ResNet18 92.2 94.8 84.1 89.3 81.2 88.3 90.2 79.6 ABCNet ResNet18 92.7 95.2 84.5 89.7 85.3 89.5 90.7 81.3 BoTNet ResNet18 89.9 92.1 81.8 88.7 71.3 84.8 88.0 74.3 BANet ResT-Lite 92.2 95.2 83.8 89.9 86.8 89.6 90.5 81.4 Segmenter ViT-Tiny 89.9 93.0 81.2 88.9 67.6 84.1 88.1 73.6 UNetFormer ResNet18 92.7 95.3 84.9 90.6 88.5 90.4 91.0 82.7 DSCT Swin-Tiny 85.9 90.7 71.8 78.6 76.2 90.2 90.3 80.6 RS3Mamba R18-Mamba-T 96.7 95.5 84.4 90.0 86.9 90.7 93.2 83.3 LASC-Mamba ResNet18 96.7 96.0 86.1 89.5 84.7 90.6 93.5 83.2 ORLA-OCM UNetFormer(本文方法) ResNet18 92.1 95.8 85.1 90.8 89.1 90.7 91.2 83.7