Polunadzirano učenje semantičke segmentacije

undergraduate thesis

• Title (english): Semi-supervised learning of semantic segmentation
• Author: Vedran Moškov
• Mentor: Siniša Šegvić (mentor)
• Committee member: Siniša Šegvić (predsjednik povjerenstva)
• Committee member: Ivan Martinović (član povjerenstva)
• Committee member: Iva Sović (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2024-07-01, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

Semantička segmentacija danas je vrlo unosno područje s mnogim primjenama, no nedostatak označenih slika je i dalje neriješen problem računalnog vida. U ovom radu isprobavamo potpuno nadzirani i polunadzirani pristup kako bismo ustanovili i ocijenili postignute točnosti. U radu su opisani osnovni koncepti dubokog učenja i metoda kojima smo se služili te korištena arhitektura. Korišteni skup podataka jest Cityscapes dok je korištena arhitektura SwiftNet temeljen na ResNet-18 arhitekturi enkodera. Programski kod je napisan u programskom jeziku Python koristeći biblioteke PyTorch, NumPy te OpenCV, dok su rezultati vizualizirani bibliotekom Matplotlib. Kod je izvršen djelomično na platformi Google Colab, a djelomično na udaljenom serveru ZEMRIS-a na FER-u.

Abstract (english)

Semantic segmentation is currently a highly lucrative field with numerous applications, yet the lack of labeled images remains an unresolved problem in computer vision. In this paper, we explore fully supervised and semi-supervised approaches to establish and evaluate the achieved accuracies. The paper describes the basic concepts of deep learning and the methods we employed, as well as the utilized architecture. The dataset used is Cityscapes, and the architecture used is SwiftNet, based on the ResNet-18 encoder architecture. The code is written in the Python programming language, utilizing libraries such as Python, NumPy and OpenCV, while the results are visualized using the Matplotlib library. The code execution was partially carried out on the Google Colab platform and partially on a remote server at ZEMRIS, FER.

• Keywords: duboko učenje; računalni vid; nadzirano učenje; polunadzirano učenje; konvolucijske neuronske mreže; semantička segmentacija
• Keywords (english): deep learning; computer vision; supervised learning; semi-supervised learning; convolutional neural networks; semantic segmentation
• Language: croatian
• URN:NBN: urn:nbn:hr:168:624513
• Study programme: Title: Computing; Study programme type: university; Study level: undergraduate; Academic / professional title: sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva (sveučilišni/a prvostupnik/ prvostupnica (baccalaureus/baccalaurea) inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2024-10-12 21:39:07