Tilapia Disease Detection Using Convolutional Neural Networks with Transfer Learning

Abstract

Disease outbreaks remain a major challenge in tilapia (Oreochromis niloticus) aquaculture, often causing significant economic losses and reduced production efficiency. Early and accurate disease detection is therefore essential to improve fish health management and support sustainable aquaculture practices. This study proposes an image-based tilapia disease detection system using Convolutional Neural Networks (CNNs) with a transfer learning approach based on the MobileNetV2 architecture. The experimental workflow consisted of image preprocessing, data augmentation, model training, and performance evaluation using accuracy, precision, recall, F1-score, and confusion matrix metrics. The proposed model was trained to classify four categories, namely Normal, Motile Aeromonas Septicemia (MAS), Parasitic Disease, and Tilapia Lake Virus (TiLV). Experimental results demonstrated that the model achieved an overall accuracy of 94%, with balanced precision, recall, and F1-score values of 94%, indicating reliable classification performance across all disease categories. Furthermore, the trained model was successfully deployed as a Streamlit-based web application, enabling automated disease prediction from uploaded fish images. These findings demonstrate that the integration of MobileNetV2 and transfer learning provides an effective and computationally efficient solution for image-based tilapia disease detection, with promising potential for supporting intelligent disease diagnosis in aquaculture.