Table 2 Comparative results of prediction on testing set.(%)
From: Self-relabeling for noise-tolerant retina vessel segmentation through label reliability estimation
Dataset | DRIVE(R) | STARE(VK) | CHASE | ||||
|---|---|---|---|---|---|---|---|
Group | Method | F1 | PR | F1 | PR | F1 | PR |
\(\mathrm {LV}\)-1 | U-Net | 73.9 | 82.9 | 79.6 | 87.4 | 77.4 | 85.8 |
Cas | 73.6 | 81.9 | 75.1 | 82.1 | 73.4 | 80.7 | |
SF | 76.3 | 84.9 | 80.4 | 88.2 | 76.5 | 84.7 | |
Ours | 76.8 | 85.5 | 81.3 | 88.9 | 77.5 | 85.3 | |
\(\mathrm {LV}\)-2 | U-Net | 74.5 | 81.3 | 79.1 | 87.3 | 70.7 | 78.4 |
Cas | 73.4 | 82.0 | 74.4 | 82.0 | 73.1 | 80.5 | |
SF | 73.7 | 82.4 | 80.0 | 87.6 | 75.3 | 82.9 | |
Ours | 75.9 | 84.8 | 80.8 | 88.1 | 75.9 | 83.4 | |
\(\mathrm {LV}\)-3 | U-Net | 72.3 | 81.4 | 78.6 | 86.4 | 70.1 | 77.0 |
Cas | 72.2 | 80.2 | 69.7 | 76.1 | 69.9 | 76.6 | |
SF | 72.0 | 80.3 | 76.9 | 85.3 | 74.4 | 81.5 | |
Ours | 75.7 | 84.3 | 79.2 | 87.0 | 74.8 | 82.8 | |
Pseudo | U-Net | 78.1 | 86.5 | 80.1 | 88.3 | / | / |
Cas | 74.0 | 82.2 | 75.8 | 83.8 | / | / | |
SF | 78.5 | 87.0 | 80.3 | 88.5 | / | / | |
Ours | 78.7 | 87.2 | 80.6 | 88.6 | / | / | |
Manual | U-Net | 80.2 | 88.7 | 80.0 | 87.9 | 77.7 | 85.0 |
Cas | 76.5 | 85.1 | 73.8 | 81.1 | 73.9 | 81.0 | |
SF | 80.9 | 89.5 | 81.3 | 88.7 | 79.4 | 87.2 | |
Ours | 81.3 | 89.8 | 82.0 | 89.1 | 80.0 | 88.0 | |