Table 2 Table presenting the numerical comparison on the Ottawa dataset, with units in percentage (%). The top-performing values are marked in bold, while those ranking second are underlined.
From: Heterogeneous dual-decoder network for road extraction in remote sensing images
Method | Source | IoU | Precision | Recall | F1 | mIoU |
|---|---|---|---|---|---|---|
D-LinkNet15 | CVPR 18 | 86.83 | 94.28 | 91.65 | 92.95 | 92.35 |
NL-LinkNet16 | GRSL 21 | 87.85 | 93.54 | 93.52 | 93.53 | 92.93 |
CoANet36 | TIP 21 | 90.36 | 95.12 | 94.75 | 94.93 | 94.40 |
DBRANet24 | GRSL 22 | 89.04 | 95.10 | 93.33 | 94.20 | 93.64 |
DDU-Net18 | TGRS 22 | 90.17 | 94.65 | 95.01 | 94.83 | 94.29 |
DSCNet20 | ICCV 23 | 88.78 | 93.16 | 94.98 | 94.06 | 93.47 |
CFRNet30 | GRSL 24 | 90.56 | 95.74 | 94.36 | 95.04 | 94.52 |
OARENet39 | TGRS 24 | 90.92 | 95.34 | 95.14 | 95.24 | 94.73 |
Swin-GAT42 | RS 25 | 91.04 | 95.79 | 94.83 | 95.31 | 94.80 |
CGCNet43 | TGRS 25 | 90.81 | 95.31 | 95.06 | 95.19 | 94.67 |
HDDNet | Ours | 91.85 | 96.20 | 95.31 | 95.75 | 95.27 |
