Development Of A CNN-Based Model For Automated Detection And Growth Monitoring Of Catfish (Clarias Gariepinus) In Aquaponics Systems

Abstract

Food security continues to face challenges from environmental degradation and climate change, and highlighting the need for innovations in sustainable aquaculture. In the Philippines, most systems remain to rely on manual fish and water parameter monitoring, which is labor intensive, not consistent, and tends to stress the fish. Although catfish (Clarias gariepinus) is a significant aquaculture species, growth monitoring is still highly manual and subject to human error. In this study, a YOLOv11 model on Convolutional Neural Networks (CNNs) was developed to facilitate the detection and monitor the growth of catfish in aquaponics systems. Images were taken employing a low-cost webcam with a resolution of 1080p within the greenhouse facility at Central Luzon State University. Images marked with annotations were subsequently run through Python and OpenCV to enhance the resolution and to extract features correctly. The performance of the model was 98% accurate for training and 96.7% for validation with a mean identification accuracy of 80% when it was trained for 50 epochs. Regression outputs indicated strong correlations (R² = 0.91 for length; R² = 0.82 for weight) with a mean error of ±55 g. Ongoing work is aimed at real-time integrating with IoT-based monitoring to facilitate smarter, more efficient, and sustainable aquaponics systems.