Figures & data
Table 1. GID dataset
Table 2. Settings of hyper-parameters
Table 3. Numerical evaluation on GID-C (%)
Figure 9. Results on GID-C test set, (a) Original image, (b) Ground truth, (c) SegNet, (d) U-Net, (e) ResUNet-a, (f) SCAttNet, (g) Ours
![Figure 9. Results on GID-C test set, (a) Original image, (b) Ground truth, (c) SegNet, (d) U-Net, (e) ResUNet-a, (f) SCAttNet, (g) Ours](/cms/asset/b02d5945-54bd-4d07-9717-182f8a9c7063/tres_a_1876272_f0009_c.jpg)
Table 4. Numerical evaluation on GID-E (%)
Figure 10. Results on GID-E test set, (a) Original image, (b) Ground truth, (c) SegNet, (d) U-Net, (e) ResUNet-a, (f) SCAttNet, (g) Ours
![Figure 10. Results on GID-E test set, (a) Original image, (b) Ground truth, (c) SegNet, (d) U-Net, (e) ResUNet-a, (f) SCAttNet, (g) Ours](/cms/asset/30f5b9af-4171-4960-b273-0f831cae1a0d/tres_a_1876272_f0010_c.jpg)
Table 5. Execution time comparison
Table 6. Training time per epoch (×103 s)
Table A1. Class-wise IoU of GID-C test set (%)
Table B1. Class-wise IoU of GID-E test set (%)