Coral Detection based on Optimised Lightweight YOLO Model

Coral reefs are essential marine ecosystems that face significant threats due to climate change, pollution, and overfishing. Effective monitoring is crucial for conservation efforts, but traditional methods are labor-intensive and inefficient. This study proposes a deep learning-based coral detection model built based on the YOLOv8 architecture, specifically for nano and small. In addition, the Ghost modules and Ghost bottlenecks were utilized to modify the original YOLOv8 small. The proposed model was trained on an underwater coral dataset and evaluated in terms of precision, recall, and mean Average Precision (mAP) metrics. Experimental results demonstrate that the YOLOv8 small model and YOLOv8 small model with Ghost modules achieved a mAP of 53.675% and 55.88%, respectively, while maintaining a compact model size. This work contributes to developing efficient and lightweight coral detection systems to support conservation efforts.