Table 3 Cross validation on DRIVE(R) and STARE datasets.(%)
From: Self-relabeling for noise-tolerant retina vessel segmentation through label reliability estimation
Dataset | DRIVE(R) | STARE(VK) | |||
|---|---|---|---|---|---|
Group | Method | F1 | PR | F1 | PR |
\(\mathrm {LV}\)-1 | U-Net | 68.5 | 75.6 | 70.3 | 78.6 |
Cas | 36.0 | 36.0 | 70.8 | 74.8 | |
SF | 65.0 | 71.2 | 72.2 | 80.2 | |
Ours | 73.9 | 81.2 | 74.5 | 82.3 | |
\(\mathrm {LV}\)-2 | U-Net | 68.7 | 74.6 | 70.1 | 77.3 |
Cas | 47.0 | 48.5 | 72.4 | 79.8 | |
SF | 66.7 | 72.6 | 70.9 | 78.3 | |
Ours | 71.8 | 79.0 | 73.0 | 79.9 | |
\(\mathrm {LV}\)-3 | U-Net | 49.0 | 46.1 | 67.0 | 74.7 |
Cas | 50.0 | 54.5 | 70.5 | 77.9 | |
SF | 67.8 | 73.5 | 68.7 | 76.4 | |
Ours | 70.2 | 76.4 | 70.5 | 78.4 | |
Pseudo | U-Net | 57.4 | 62.7 | 72.2 | 78.1 |
Cas | 49.6 | 52.2 | 76.9 | 83.5 | |
SF | 70.9 | 76.9 | 74.7 | 81.1 | |
Ours | 76.3 | 84.4 | 77.8 | 85.7 | |
Manual | U-Net | 55.3 | 59.7 | 70.6 | 77.3 |
Cas | 61.9 | 68.3 | 75.5 | 82.0 | |
SF | 65.9 | 71.3 | 74.9 | 82.7 | |
Ours | 71.4 | 78.6 | 77.1 | 83.6 | |