Skip to main content
. 2023 Jul 11;9:e1467. doi: 10.7717/peerj-cs.1467

Table 4. Quantitative results on the public dataset CITY-OSM.

Method Backbone Background (%) Building (%) Road (%) OA (%) Mean F1 (%) mIoU (%)
DC-Swin 63.71 68.79 55.06 85.42 76.79 62.52
SegFormer 67.35 72.69 61.73 87.48 80.34 67.26
GCNet ResNet18 66.67 71.76 63.44 87.32 80.39 67.28
A2FPN ResNet18 71.12 75.72 77.65 89.26 83.34 71.50
PSPNet ResNet18 75.40 79.01 73.10 91.06 86.24 76.04
BuildFormer 75.36 79.30 72.79 91.08 86.22 75.82
LOANet (ours) ResNet18 73.17 76.54 70.76 90.04 84.70 73.49
LOANet (ours) LDCNet (ours) 73.95 78.25 70.95 90.49 85.28 74.39
LOANet-Large (ours) LDCNet-large (ours) 75.51 79.51 73.30 91.17 86.41 76.10

Notes.

Bold values represent the best performance indicators.