Vícetřídá segmentace 3D skenů magnetické rezonance pomocí strojového učení
Loading...
Date
2020-04-30
Authors
ORCID
Advisor
Referee
Mark
Journal Title
Journal ISSN
Volume Title
Publisher
International Society for Science and Engineering, o.s.
Abstract
Následující práce se zabývá segmentací obrazu do více tříd pomocí konvolučních neuronových sítí. Text shrnuje úvod do problematiky segmentace obrazu, kde jsou popsány základní principy fungování neuronových sítí a je přiblížen současný vývoj strojového učení. Pro segmentaci obrazu je zvolena a podrobněji popsána architektura U-Net, která byla aplikována na dataset medicínských dat. Dále je tu popsán postup zpracování trojrozměrných dat, jejich předzpracování a metody, jakými byla provedena celá vícetřídá segmentace. Práce srovnává dvojici metod, jakými lze segmentaci provést. Vybrané metody podrobněji vyhodnocuje a zasazuje do širšího kontextu.
This article deals with multiclass image segmentation using convolutional neural networks. The theoretical part focuses on image segmentation and current machine learning development. There are basic principles of neural networks and image segmentation with more types of approaches. In practical part the U-Net architecture is chosen and is described for image segmentation more. U-Net was applied for medicine dataset. There is processing procedure which is more described for image processing of three-dimensional data. There are also methods for data preprocessing which were applied for image multiclass segmentation. Final part of this work evaluates results.
This article deals with multiclass image segmentation using convolutional neural networks. The theoretical part focuses on image segmentation and current machine learning development. There are basic principles of neural networks and image segmentation with more types of approaches. In practical part the U-Net architecture is chosen and is described for image segmentation more. U-Net was applied for medicine dataset. There is processing procedure which is more described for image processing of three-dimensional data. There are also methods for data preprocessing which were applied for image multiclass segmentation. Final part of this work evaluates results.
Description
Keywords
Citation
Document type
Peer-reviewed
Document version
Published version
Date of access to the full text
Language of document
cs
Study field
Comittee
Date of acceptance
Defence
Result of defence
Document licence
(C) 2020 Elektrorevue