Coral Reef Disease and Bleaching Indentification through Computational Vision Algorithm

Degradation in coral reefs has increased at a worrying rate because of human activity such as overfishing, sedimentation and tourism; triggering climate change and other factors that affect the waters where they reside. All of this causes a great source of stress that weakens the coral’s immune system. The characteristics of coral diseases were studied, as well as how to identify a healthy coral and one that suffers from bleaching using recognition and the use of convolutional neural networks to determine the most distinctive characteristics of these diseases, it was possible to develop the algorithm, using the YOLO artificial intelligence infrastructure for the training process of a computer vision algorithm. For this research the incremental methodology was developed in 3 phases of training for evaluate the accuracy, modifying different parameters. After three increments using 1,555 images the average loss values remained at recommended values until reaching a value of 0.3427, on the other hand, the mAP decreased to a value of 71.35%. The last validation process showed a satisfactory result with an average accuracy of 94.12% in the detection of SCTLD, the most prevalent disease in recent years in the Caribbean.