From: Scale-aware dense residual retinal vessel segmentation network with multi-output weighted loss
 | DRIVE dataset | STARE dataset | ||||||
---|---|---|---|---|---|---|---|---|
network | dice | acc | mIoU | recall | dice | acc | mIoU | recall |
UNet [4] | 0.7751 | 0.9622 | 0.7963 | 0.8661 | 0.8103 | 0.9705 | 0.8251 | 0.8693 |
Attention U-Net [10] | 0.7849 | 0.9644 | 0.8044 | 0.8720 | 0.8158 | 0.9718 | 0.8297 | 0.8769 |
CE-Net [13] | 0.7794 | 0.9630 | 0.7996 | 0.8678 | 0.8141 | 0.9711 | 0.8280 | 0.8737 |
SA-Unet [5] | 0.7993 | 0.9654 | 0.8120 | 0.8783 | 0.8284 | 0.9731 | 0.8395 | 0.8892 |
Sine-Net [22] | 0.8006 | 0.9665 | 0.8167 | 0.8757 | 0.8303 | 0.9738 | 0.8414 | 0.8858 |
DR-Vnet [9] | 0.8011 | 0.9665 | 0.8165 | 0.8782 | 0.8307 | 0.9733 | 0.8419 | 0.8844 |
CRAUnet [21] | 0.8028 | 0.9681 | 0.8186 | 0.8763 | 0.8325 | 0.9746 | 0.8441 | 0.8855 |
Our network | 0.8040 | 0.9667 | 0.8214 | 0.8810 | 0.8341 | 0.9739 | 0.8438 | 0.8884 |