DSpace Repository

Deep Instance Segmentation and Polygonization

Show simple item record

dc.contributor.advisor Dailey, Matthew
dc.contributor.author Deshapriya, N Lakmal
dc.contributor.other Miyazaki, Hiroyuki
dc.contributor.other Hazarika, Manzul
dc.date.accessioned 2020-05-13T10:25:20Z
dc.date.available 2020-05-13T10:25:20Z
dc.date.issued 2020-05
dc.identifier.other AIT
dc.identifier.uri http://www.cs.ait.ac.th/xmlui/handle/123456789/970
dc.description 57 p. en_US
dc.description.abstract Current advances in deep learning aspects of machine learning are leading to new results on computer vision tasks such as object classification, localization, semantic segmentation, and instance segmentation. Machine learning systems are now achieving human-level accuracy in many tests. In this study, I develop a new deep learning technique (deep convolutional neural network architecture) for instance segmentation tasks. Each instance is approximated by a polygon with a finite number of edges (polygonization) to produce a GIS shapefile. I demonstrate the feasibility of the proposed architecture with respect to instance segmentation tasks on satellite images, which have a wide range of applications. Moreover, I demonstrate the usefulness of the new method for extracting building foot-prints from satellite images. Total pixel-wise accuracy of my approach was 89 % reaching close to accuracy of state-of-the-art Mask RCNN approach (91 %). And my approach provides alternative approach to instance segmentation with a simpler and more intuitive neural network. en_US
dc.description.sponsorship Asian Institute of Technology (AIT) en_US
dc.publisher AIT en_US
dc.subject Deep Learning en_US
dc.subject Instance Segmentation en_US
dc.subject Parametrization en_US
dc.subject Satellite Images en_US
dc.title Deep Instance Segmentation and Polygonization en_US
dc.type Thesis en_US

Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace

Advanced Search


My Account