1Research Scholar, Department of Computer Science and Engineering, Annamalai University, Chidambaram, India
2Associate Professor, Department of Computer Science and Engineering, Annamalai University, Chidambaram, India
3Associate Professor, Department of Computer Science and Engineering, Sree Dattha Group of Institutions, Telangana, India
Coral reef ecosystems are very important maritime environments, which are experiencing degradation effects due to climate change. Coral reef detection is very essential, though challenging, given the changing dynamics while coral ecosystem data collection processes are sensitive to weather changes. At most times, there have been difficulties in collecting precise data. For coral reef surveys, the processes have been conducted during optimal conditions, which are likely to be absent when individuals are on-site. This is just a small part of the complicated relationships in coral reef modeling. Thus, it is imperative to explore the possibility of using deep learning in classifying coral reefs. This investigation results in exploring the possibility of using non-ideal shoreline photographs for classifying coral reefs. This would give room to increase the scope of other approaches of studying coral reef surveys. For the images in the Coral Reef data set, there is categorization using a segmentation method, which involves a convolution neural network. The categorifications have advanced high accuracy of 92% using MobileNetv2 (Transfer Learning), 89% using Quad-Layer CNN Architecture, while the other portion of 88% using Tri-Layer CNN Architecture. There have been difficulties in classifying the non-ideal images, though with high chances of precise results.
Keywords: Machine Learning, Deep Learning, marine images, corals, classification, convolutional neural networks.
How to cite this article: Parveen MZ, Sudha LR, Santhoshkumar R. Transfer learning–based coral classification using convolutional neural networks. Int J Drug Deliv Technol. 2026;16(8s): 59-64; DOI: 10.25258/ijddt.16.8s.12
Source of support: None.
Conflict of interest: None