Image processing and al applied to weldings

Papachlimintzos, Ioannis; Παπαχλιμίντζος Ιωάννης

dc.contributor.author	Papachlimintzos, Ioannis	en
dc.contributor.author	Παπαχλιμίντζος Ιωάννης	el
dc.date.accessioned	2023-02-01T11:55:41Z
dc.date.available	2023-02-01T11:55:41Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/57025
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.24723
dc.rights	Αναφορά Δημιουργού-Μη Εμπορική Χρήση-Όχι Παράγωγα Έργα 3.0 Ελλάδα	*
dc.rights.uri	http://creativecommons.org/licenses/by-nc-nd/3.0/gr/	*
dc.subject	Convolutional Neural Network	en
dc.subject	Unet	en
dc.subject	Weld seams	en
dc.subject	Semantic segmentation	en
dc.subject	Deep learning	en
dc.subject	Ραφές συγκόλλησης	el
dc.subject	Συνελικτικό νευρωνικό δίκτυο	el
dc.subject	Σημασιολογική τμηματοποίηση	el
dc.subject	Βαθιά μάθηση	el
dc.title	Image processing and al applied to weldings	en
heal.type	bachelorThesis
heal.classification	Computer Science	en
heal.classification	Classification	en
heal.classification	Neural Networks	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2022-11-07
heal.abstract	Marine automated inspections using Unmanned Areal Vehicles (UAVs), Remotely Operated Vehicles (ROVs) are emerging technologies which are constantly gaining ground. Intelligent vehicles is essential to have an environmental perception that provides crucial information about each ship’s feature so then to classify and inspect it according to the regulations. In this thesis, the main task is research about the application of image segmentation in welding joints. A Fully Convolution Neural Network is proposed based on UNet architecture which was modified so that VGG16 to be implemented as an encoder following a couple of transfer learning strategies. Decoder's convolutional layers were reduced by replacing one layer on each block with Batch Normalization and Dropout operations in order to minimize computational cost and increase model's accuracy. The dataset used for the training and testing of the model consists of images with welding joints which were collected from school's laboratory, ship surveys as well as from the internet (300 images) to achieve greater diversity and increase model’s robustness. The experimental results of the testing set show that the mean IoU is 0.46 and mean F1-score is 0.60.	en
heal.advisorName	Papalambrou, George	en
heal.advisorName	Samuelides, Manolis	en
heal.committeeMemberName	Zervaki, Anna	en
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Ναυπηγών Μηχανολόγων Μηχανικών. Τομέας Ναυτικής Μηχανολογίας. Εργαστήριο Ναυτικής Μηχανολογίας	el
heal.academicPublisherID	ntua
heal.numberOfPages	71 σ.	el
heal.fullTextAvailability	false