Coral Reef Image Classification Using Multilayer Perceptron

Abstract

Coral reefs are one of the marine organisms that play many crucial roles for other organisms within them. Coral reefs are often referred to as tropical rainforests because they serve as shelters for small fish and produce food for other marine organisms. Over time, various threats have emerged that disrupt the stability of the marine ecosystem, one of which is coral reef degradation, such as bleaching or physical damage caused by multiple factors. These factors include climate change, chemicals resulting from fishing with explosives, and pollution. As a result, coral reefs become damaged and can no longer serve as a refuge for small species.  Therefore, this study aims to mitigate the impact of coral reef damage by developing a coral reef classification model using one of the deep learning algorithms and artificial neural networks, namely the Multilayer Perceptron (MLP), which employs multiple hidden layers in its modeling process. The classification results using this algorithm achieved an accuracy of 73%, indicating that the model performs well in classifying coral reefs in image form. Thus, it is hoped that deep learning innovations for coral reef classification can contribute significantly to coral reef conservation and marine resource management.