Coral Detection with Machine Learning

Ocean exploration has always been a costly, high-risk, and slow process, but it is an essential part of the environment. In order to reduce dependence on human labor and increase the speed of data processing, we will introduce a method of using machine learning to assist in coral area surveys in this paper. We will first provide an introduction to the principles of coral image recognition and the structure of the SOLOv2 model, followed by a presentation of the method used to train the model to identify corals, the results of further optimizing the original SOLOv2 model, and an algorithm used to estimate area.