Zero-shot capabilities of advanced image segmentation models

Kastoris, Charalampos; Καστόρης, Χαράλαμπος

dc.contributor.author	Kastoris, Charalampos	en
dc.contributor.author	Καστόρης, Χαράλαμπος	el
dc.date.accessioned	2025-01-08T12:24:39Z
dc.date.available	2025-01-08T12:24:39Z
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/60660
dc.identifier.uri	http://dx.doi.org/10.26240/heal.ntua.28356
dc.rights	Default License
dc.subject	SAM	en
dc.subject	Segmentation	en
dc.subject	CLIPSeg	en
dc.title	Zero-shot capabilities of advanced image segmentation models	en
heal.type	bachelorThesis
heal.secondaryTitle	Pushing SAM and CLIP to their limits	en
heal.classification	Machine learning	en
heal.language	en
heal.access	free
heal.recordProvider	ntua	el
heal.publicationDate	2024-07-09
heal.abstract	The object of this work is to study image segmentation across various reference datasets, utilizing two state-of-the-art deep learning models, Segment Anything Model (SAM) and CLIPSeg. The aim is to explore the performance of these models for the task of image segmentation across multiple datasets, employing metrics such as Intersection over Union (IoU) and Dice Loss. Image segmentation has become a critical task in computer vision due to the exponential growth of image data volume and has numerous applica tions in various domains. Image segmentation involves dividing an image into multiple segments, where each segment represents a different object or region within the image.	en
heal.advisorName	Βουλοδήμος, Αθανάσιος	el
heal.committeeMemberName	Στάμου, Γεώργιος	el
heal.committeeMemberName	Σταφυλοπάτης, Ανδρέας	el
heal.academicPublisher	Εθνικό Μετσόβιο Πολυτεχνείο. Σχολή Ηλεκτρολόγων Μηχανικών και Μηχανικών Υπολογιστών. Τομέας Τεχνολογίας Πληροφορικής και Υπολογιστών	el
heal.academicPublisherID	ntua
heal.numberOfPages	131 σ.	el
heal.fullTextAvailability	false