Curvilinear Structure Segmentation on Private Retinal Dataset (test)
76.99Dice CoefficientTopology SegNet
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Topology SegNetLoss=CE+Dice Loss, Feature=PI Integration2026.01 | 76.99 | 82.77 | 87.72 | 4.278 | 0.464 | |
| Topology SegNetLoss=cbDice, Feature=PI Integration2026.01 | 76.43 | 81.67 | 87.14 | 4.41 | 7.289 | |
| Topology SegNetLoss=clDice, Feature=PI Integration2026.01 | 76.15 | 82.7 | 86.99 | 7.049 | 0.537 | |
| Topology SegNetLoss=CE Loss, Feature=PI Integration2026.01 | 75.78 | 83.15 | 87.64 | 5.821 | 0.629 | |
| Topology SegNetLoss=PH Loss, Feature=PI Integration2026.01 | 75.54 | 82.65 | 87.52 | 6.869 | 0.534 | |
| cbDiceLoss=cbDice2026.01 | 75.15 | 81.08 | 86.8 | 8.503 | 6.01 | |
| clDiceLoss=clDice2026.01 | 74.89 | 81.85 | 86.97 | 7.906 | 0.587 | |
| U-NetLoss=CE Loss2026.01 | 74.84 | 81.69 | 86.47 | 7.765 | 0.582 | |
| PH LossLoss=PH Loss2026.01 | 74.58 | 81.36 | 86.74 | 7.083 | 0.549 |