Deep or Shallow Networks: Coral Types Classification Using Residual Network Models

Citation

Nurill-Nabilla, H. and Mohd Isa, Wan Noorshahida (2024) Deep or Shallow Networks: Coral Types Classification Using Residual Network Models. Springer Nature Link, 1183. pp. 77-89. ISSN 1876-1100

Full text not available from this repository.

Abstract

Coral reefs help to maintain healthy shorelines and support the growth of mangroves and other coastal vegetation as well as becoming home to many diverse types of sea creatures. Aside from benefitting marine life, they also benefit humans by providing critical coastal protection by reducing the impact of storms and erosion. Due to anthropogenic threats, coral reef habitats are on its path to destruction. Marine scientists are currently working on ways to restore and rehabilitate coral reefs for the good of all organisms that benefit from coral reefs. These efforts by the marine scientists take a significant amount of time to execute. Thus, marine biologists and computer vision experts work together to enhance their methods of coral conservation and rehabilitation with the help of machine learning and autonomous underwater vehicles (AUV). In this paper, we propose an experiment to teach small datasets of texture corals to pre-trained deep residual networks for them to be able to classify these corals. In doing this, the machine will be able to assist humans in the maintenance and preservation of coral reef ecosystems. The models that will be used in this project are ResNet18, ResNet50, and ResNet152. These models are chosen due to their performance in recent studies in the field of coral recognition and classification. In this paper, we have achieved 96.08% as the highest accuracy of ResNet18, 98.87% for ResNet50 and 98.37% for ResNet152.

Item Type: Article
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK5101-6720 Telecommunication. Including telegraphy, telephone, radio, radar, television
Divisions: Faculty of Computing and Informatics (FCI)
Depositing User: Ms Nurul Iqtiani Ahmad
Date Deposited: 04 Nov 2024 01:34
Last Modified: 04 Nov 2024 01:34
URII: http://shdl.mmu.edu.my/id/eprint/13083

Downloads

Downloads per month over past year

View ItemEdit (login required)