Traditional diagnosis methods for detecting aeromonas hydrophila disease often require time and effort, and the results are not always consistent. This leads to delays in disease management, resulting in economic losses. In addition, consumption of contaminated fish can also pose health risks to humans, such as gastroenteritis, skin infections, and sepsis. This research aims to develop a diagnosis method for Aeromonas hydrophila disease in freshwater fish using the Convolutional Neural Network (CNN) method. Aeromonas hydrophila disease, which is often caused by poor environmental conditions, has significantly impacted the ecosystem and economy of the fisheries sector in Indonesia, with losses reaching 12 trillion rupiah per year. The study utilises the MobileNetV2 model for classification of three categories of fish images: healthy, infected with aeromonas hydrophila, and other diseases (white spots). The data used consisted of fish images processed with augmentation techniques to increase dataset variation. The model was trained using transfer learning with InceptionV3 architecture, combined with a fully connected layer and softmax activation function for classification. Performance evaluation was conducted using accuracy (80%), precision (86%), recall (80%), F1-score (78%) metrics, as well as confusion matrix analysis. Results showed the model was able to achieve 96.57% accuracy on test data, with stable performance on validation data. The results of this research show the great potential of CNN in detecting fish diseases automatically and efficiently. The research results are expected to contribute to improving the quality of fish farming, food safety, and economic resilience of the fisheries sector.