An Intelligent Fish Feeder System Based on Biomass Information Using Computer Vision and IoT Integration

Small-scale tilapia aquaculture is constrained by labor shortages, inefficient feed management, and difficulties in maintaining stable water quality, all of which limit productivity and sustainability. Rising feeding costs account for over 50 percent of operational expenses, while disease outbreaks linked to limited genetic acclimatization exacerbate stress and growth issues, further threatening industry profitability. Existing "intelligent" feeding systems often focus on either environmental monitoring or image-based biomass estimation, but rarely integrate both in a low-cost, real-time solution suitable for smallholders. To address these constraints, this study introduces an intelligent feeding system that integrates Internet of Things (IoT) sensors for water quality monitoring with a You Look Only Once (YOLO)v8-based computer vision model for real-time biomass estimation. The system automatically adjusts feeding based on live video analysis and environmental data. The YOLOv8 model achieved a 42.2 percent improvement in precision and a 9.2 percent increase in mean average precision (mAP) over two development stages. The IoT water quality monitoring system (WQMS) demonstrated 98 percent accuracy compared to benchmark data from the Yellow Springs Instrument (YSI) Professional Plus. Real-time deployment at the International Institute of Aquaculture and Aquatic Sciences (I-AQUAS), Universiti Putra Malaysia, validated the system's capability to reduce feed wastage and maintain water quality. By addressing the dual challenges of feed inefficiency and water quality degradation, this integrated solution enhances farm management efficiency, reduces costs, and strengthens the sustainability and competitiveness of small-scale tilapia aquaculture.