Real-time rapid visual fish disease detection system based on tiny machine learning

Abstract

The rapid detection of fish diseases is crucial for the sustainable development of aquaculture, ensuring both economic viability and environmental protection. This study presents a novel real-time fish disease detection system based on tiny machine learning (TinyML) technology. By integrating the You Only Look Once 11 nano (YOLO11n) lightweight object detection model with a RISC-V microcontroller and hardware design, the system achieves efficient, low-power, and accurate disease detection tailored to resource-constrained aquaculture environments. The system incorporates edge computing to perform real-time disease detection locally, reducing reliance on cloud services and improving data security. Experimental results demonstrate the system’s effectiveness, achieving a mean average precision at IoU thresholds from 0.5 to 0.95 (mAP50–95) of 0.736 with robust performance in real-world scenarios. The lightweight architecture enables flexible deployment in various aquaculture conditions, from offshore environments to small-scale farms. This study underscores the potential of TinyML to revolutionize aquaculture management and promote the intelligent, automated monitoring of fish health.