NORMA eResearch @NCI Library

Marine Debris Segmentation using Capsule Network (SegCaps)

Shinde, Palash (2021) Marine Debris Segmentation using Capsule Network (SegCaps). Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
Preview
PDF (Master of Science)
Download (572kB) | Preview
[thumbnail of Configuration manual]
Preview
PDF (Configuration manual)
Download (307kB) | Preview

Abstract

Disposal of ever-increasing marine debris is proving to be one of the difficult tasks to perform. Making the environment more sustainable requires the correct recycling and disposal of this ever-increasing garbage. Ocean bodies are being inundated every day by millions of plastic garbage pieces. This is having a significant impact on the aquatic life system. When debris is identified and classified by kind on an instance level in an automated manner, debris can be tracked on a large scale without involving humans. CNN models such as Mask R-CNN, Faster R-CNN is often regarded as a cutting-edge approach for object segmentation in an image, its architectures come with certain limitations. To tackle, the limitations of traditional CNN Networks, this research aims to introduce a new Capsule network-based objection segmentation technique called as SegCaps that has the capabilities to tackle the limitations of traditional CNN-based segmentation techniques in terms of generating accurate object segmentation, while at the same time reducing the network's complexity in terms of depths layer. The SegCaps model introduced in the research is trained on the Trashcan dataset, which has 7,212 object images of 3 main categories, annotated on mask level. These main categories are further divided into 16 subcategories classes. The model in the proposed research achieved an overall mean average precision of (mAP) of 26.25 & dice coefficient of 28.46 with an IOU threshold of 0.5 for the segmentation task.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Instance Segmentation; Marine Debris; Capsule Networks
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
Q Science > QA Mathematics > Computer software
T Technology > T Technology (General) > Information Technology > Computer software
G Geography. Anthropology. Recreation > GE Environmental Sciences > Environment
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Clara Chan
Date Deposited: 14 Dec 2021 13:07
Last Modified: 14 Dec 2021 13:07
URI: https://norma.ncirl.ie/id/eprint/5220

Actions (login required)

View Item View Item