Automatic Coral Morphotypes Detection with YOLO: A Deep Learning Approach for Efficient and Accurate Coral Reef Monitoring
Abstract
Coral reefs harbor a large portion of marine biodiversity but are declining rapidly. Conservation efforts rely on monitoring coral abundance and composition as predominant indicators of ecosystem health and management success. However, manual monitoring of coral abundance proves arduous where artificial intelligence-based automatization can help improve efficiency and accuracy. This paper presents a methodology using YOLOv5-based deep learning for automatic detection of corals and classification by morphotype, representing an important step toward streamlining machine-assisted coral reef monitoring. The research addresses the escalating need for precise and timely ecosystem assessments amidst increasing ecological shifts of coral reefs. Using state-of-the-art object detection techniques, the study strives to streamline the detection and classification of diverse coral morphotypes, which is essential for understanding reef dynamics and assessing conservation efforts. To train and evaluate our system, we use a dataset consisting of 280 original underwater coral reef images. We increased the number of annotated images to 388 by manipulating images using data augmentation techniques, which can improve model performance by providing more diverse examples for training. Our system leverages the YOLOv5 algorithm’s real-time object detection capabilities, enabling efficient and accurate coral detection. We used YOLOv5 to extract discriminating features from the annotated dataset, enabling the system to generalize, including previously unseen underwater images. The successful implementation of the automatic coral morphotypes detection system with YOLOv5 on our original image dataset highlights the potential of advanced computer vision techniques for coral reef research and conservation.