Ensemble learning approaches for retinal vessel segmentation

Abstract

Retinal vessel analysis of fundus images is an important practice for the screening and diagnosis of related diseases. Yet, automatic segmentation remains a challenging task. It is well known that ensemble learning methods show great effectiveness improving models performance in a number of applications. Bearing this in mind, in this paper, we explore the implementation of two ensemble techniques, Stochastic Weight Averaging and Snapshot Ensembles, for retinal vessel segmentation. The proposed methods are verified on DRIVE database and it shows higher performance, in terms of Acc, when compared with other state-of-the-art methods. Also, our results hint that may be possible to further improve the segmentation performance, tuning these ensemble methods.