Nadzorovano globoko učenje za segmentacijo razpok v betonu

diplomsko delo

Matic Šuc (Author), Danijel Skočaj (Mentor), Domen Tabernik (Co-mentor)

Abstract

Eden izmed problemov upravljanja z infrastrukturo je pregledovanje njene kakovosti, ki preverja stanje infrastrukture kot so cestišča, mostovi in podobni objekti. Razpoke so zelo zgodnji indikator morebitnega slabšanja stanja infrastrukture objektov, kar je lahko nevarno za uporabnike. Hitra in natančna detekcija razpok lahko bistveno zmanjša stroške vzdrževanja in izboljša učinkovitost. V diplomski nalogi je predstavljena rešitev tega problema z nadzorovanim globokim učenjem za segmentacijo razpok v betonu. Predstavljeni so tudi dodatki k rešitvi, ki znatno pripomorejo k izboljšanju učinkovitosti in zmogljivosti modela. Rešitev je ovrednotena na več različnih slikovnih množicah ter primerjana s sorodnimi pristopi.

Keywords

nevronske mreže;segmentacija;klasifikacija;razpoke v betonu;visokošolski strokovni študij;diplomske naloge;

Data

Language:	Slovenian
Year of publishing:	2022
Typology:	2.11 - Undergraduate Thesis
Organization:	UL FRI - Faculty of Computer and Information Science
Publisher:	[M. Šuc]
UDC:	004.8:625.821.5(043.2)
COBISS:	121455363
Views:	39
Downloads:	16
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Supervised deep learning for concrete crack segmentation
Secondary abstract:	One of the problems of infrastructure maintenance is the review of its quality, which controls the state of infrastructure, such as roads, bridges and similar objects. Cracks are a very early indicator of the possible deterioration of infrastructure objects, which can be dangerous for users. Fast and accurate detection of these cracks can reduce maintenance costs and improve efficiency. The diploma thesis presents a solution to this problem by applying supervised deep learning for detection of cracks on concrete surfaces. Additions to the solution are also presented, which significantly help to improve the efficiency and performance of the model. The solution was tested on several different image datasets and compared to related approaches.
Secondary keywords:	neural networks;segmentation;classification;computer vision;deep learning;concrete cracks;computer science;diploma;Globoko učenje (strojno učenje);Računalniški vid;Beton;Računalništvo;Univerzitetna in visokošolska dela;
Type (COBISS):	Bachelor thesis/paper
Study programme:	1000470
Embargo end date (OpenAIRE):	1970-01-01
Thesis comment:	Univ. v Ljubljani, Fak. za računalništvo in informatiko
Pages:	51 str.
ID:	16391441