Deep instance segmentation and polygonization | |
| Author | Deshapriya, Nawarathnage Lakmal |
| Subject(s) | Machine learning Image segmentation Neural networks (Computer science) |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Computer Science |
| Publisher | Asian Institute of Technology |
| 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. |
| Year | 2020 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Dailey, Matthew N.; |
| Examination Committee(s) | Miyazaki, Hiroyuki;Hazarika, Manzul Kumar; |
| Scholarship Donor(s) | Asian Institute of Technology (AIT), Thailand; |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2020 |