Abstract

This paper demonstrates the efficacy of a modified U-Net structure for theextraction of vascular tree masks for human fundus photographs. On limitedcompute resources and training data, the proposed model only slightlyunderperforms when compared to state of the art methods.

 

Quick Read (beta)