A convolutional network for semantic segmentation with more paths for information flow. We demonstrate that LadderNet can be viewed as an ensemble of FCNs, and validate its superior performance on blood vessel segmentation task in retinal images.
A novel preprocessing flow which dramatically improves the performance of U-Net in segmenting the blood vessels from the fundus images. The features as learned by the U-Net show accurate representation of the blood vessels.
The overall performance of our method and other state-of-the-art deep-learning based methods on DRIVE and CHASE DB1 are tabulated in this paper. Table 2 for DRIVE, 3 for CHASEDB1 respectively represents the performance comparison. The results prove that our method has displayed superior performance compared to recent state-of-the-art methods.
