Fundus Retinal Blood Vessel Segmentation Based on Active Learning

Abstract

In this article, we conducted a blood vessel segmentation experiment by using an active learning method. Using fewer manually labeled pictures to train a neural network, the accuracy of the obtained blood vessel segmentation exceeded that of supervised learning. The highest accuracy is 96.97%. It is proved that active learning can reduce the workload of human annotation data and improve the accuracy of blood vessel segmentation.

Topics & Concepts