Image segmentation using deep learning

Abstract

The goal of this dissertation is to gain a thorough understanding of the field of image semantic segmentation as well as a deep dive into the domain of deep learning and how we can use various networks, particularly convolutional neural networks and, more specifically, the U-Net architecture and its variants Attention U-Net and Attention Residual U-Net, in the medical imaging field with electron microscopy images. The tests and valuation of the mentioned architectures demonstrate the advantages of using such networks and their fectiveness in solving image semantic segmentation problems