Table 4 Ablation experiments conducted on the DRIVE dataset.
From: Retina image segmentation using the three-path Unet model
No. | Model | Unet | SegNet | HarrNet | AE | DSL | Dice | SE | SP | Acc |
|---|---|---|---|---|---|---|---|---|---|---|
1 | Unet10 | \(\surd \) | Â | Â | Â | Â | 0.8157 | 0.7942 | 0.9826 | 0.9622 |
2 | SegNet21 | Â | \(\surd \) | Â | Â | Â | 0.8172 | 0.7893 | 0.9792 | 0.9550 |
3 | HaarNet | Â | Â | \(\surd \) | Â | Â | 0.8178 | 0.8013 | 0.9769 | 0.9545 |
4 | Unet+SegNet | \(\surd \) | \(\surd \) | Â | Â | Â | 0.8187 | 0.8010 | 0.9773 | 0.9548 |
5 | SegNet+HaarNet | Â | \(\surd \) | \(\surd \) | Â | Â | 0.8242 | 0.7875 | 0.9820 | 0.9565 |
6 | Unet+HaarNet | \(\surd \) | Â | \(\surd \) | Â | Â | 0.8237 | 0.8010 | 0.9790 | 0.9564 |
7 | TP-Unet | \(\surd \) | \(\surd \) | \(\surd \) | Â | Â | 0.8252 | 0.7985 | 0.9801 | 0.9569 |
8 | TP-Unet+AE | \(\surd \) | \(\surd \) | \(\surd \) | \(\surd \) | Â | 0.8256 | 0.7986 | 0.9801 | 0.9571 |
9 | TP-Unet+AE+DSL | \(\surd \) | \(\surd \) | \(\surd \) | \(\surd \) | \(\surd \) | 0.8291 | 0.8184 | 0.9773 | 0.9571 |
