Abstract
Introduction Coral reefs are among the most vital and diverse ecosystems on the planet, providing habitats for marine life, supporting fisheries, and protecting coastlines. However, they are increasingly threatened by outbreaks of Crown-of-Thorns Starfish (COTS), a coral-eating predator capable of causing large-scale reef destruction. Traditional monitoring methods rely on manual diver surveys, which are time-consuming, labour-intensive, and unsuitable for rapid large-scale assessments. Methods To address these limitations, this study proposes an AI-powered framework for detecting COTS in underwater imagery. The system integrates advanced deep learning object detection techniques with synthetic data augmentation to improve model robustness and adaptability under complex underwater conditions. Synthetic training images were generated to expand dataset variability, while optimized detection models were designed for high accuracy and real-time inference. Results The final detection model demonstrated strong performance, achieving a precision of 0.927, recall of 0.903, and mAP@50 of 0.938. These results indicate the effectiveness of the framework in accurately identifying COTS across diverse underwater environments. Discussion and Conclusion The proposed solution is designed for deployment on embedded systems, ensuring practical, scalable, and efficient monitoring of coral reef ecosystems. By enabling real-time and high-accuracy detection of COTS, this framework supports timely interventions and contributes to the conservation and ecological resilience of coral reefs, particularly in vulnerable regions such as the Great Barrier Reef.