Volume 18, No. 6, 2021
A Novel Approach For Coral Reef Disease Detection And Classification Using Deep Learning Techniques
M. H. Ibrahim , Dr. M. Mohamed Sathik
Abstract
Coral reefs are a significant and inevitable feature of the maritime environment. It has a significant impact on maintaining the ecosystem's equilibrium. For fish like tuna, dolphins, and other aquatic animals, coral reefs provide nutrition and protein. But the overall number of coral reefs has severely decreased as a result of bleaching, climate change, human activity, and industrial pollution. A hybrid ZeNet and VGG19 deep convolution neural network machine learning approach is proposed in this study effort to solve this concerning problem by identifying diseased or infected corals from movies. In the majority of classification tasks nowadays, deep neural network (DNN) models have supplanted the formerly manual feature extractors. Because of their domain independence and extensive dataset training, DNNs like ResNet, DenseNet, VGGNet, and Inceptions models provide unparalleled performance across a wide range of applications. The proposed framework extracts feature with a hybrid DCNN approach AlexNet and ZeNet and VGG19 used to further invariance that improves classification accuracy. Extensive evaluation has enabled the identification of the optimal patch, cluster size, kernel combination, and best-performing classifier.
Pages: 10029-10042
Keywords: Deep learning; Coral reefs; Corel disease; marine ecosystem; CNN, ZeNet, VGG19, Hybrid DCNN.