Figures & data
Figure 1. SegNet architecture
Figure 2. DUpsampling operation
Figure 3. Overview of proposed framework
Figure 4. Details of SLSAM
Figure 5. Details of DLCAM
Figure 6. Illustration of focal loss
Table 1. GID dataset
Figure 7. Examples of GID-C dataset, (a) Original image, (b) Ground truth
Figure 8. Examples of GID-E dataset, (a) Original image, (b) Ground truth
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
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 11. Training loss with different architecture
Figure 12. Training accuracy with different loss function
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 (%)