CoralClassify: Advancing Coral Health Monitoring using Deep Learning

Coral reefs play a vital role to enhance marine ecosystem but changes in environment and bleaching events causes danger to it. Monitoring coral health is crucial for effective condervation efforts. Therefore, an advocate solution is required to monitor coral health. In the present paper, a novel approach “CoralClassify”, a deep learning framework is used to monitor the health of coral reefs. It helps to identify the healthy and bleached corals through image classification. The proposed methodology involves the dataset preparation and preprocessing that includes the collection of images using Flickr API and images from StructureRSMAS and ReefBase standard datasets. An image augmentation algorithm is also proposed to maintain the balance of both the classes of coral reefs and helps to enhance the recognition accuracy of CoralClassify framework. By employing modified Resnet50, a deep learning model is proposed that is capable of distinguishing healthy and bleached corals from underwater images. The experimental results illustrates the supremacy of the proposed work and achieved 87.64% accuracy, precision of 87.99%, recall of 87.06% and F1-score of 87.37%. CoralClassify can help us to monitor large areas of coral reefs and take action quickly to protect them. This research demonstrates that how Deep Learning can be used to monitor the environment and help protect our oceans.