Mendeteksi Penyakit Ikan Air Tawar dengan Metode Convolutional Neural Network (CNN)

Abstract

Diseases in freshwater fish are one of the main factors contributing to reduced survival rates, decreased aquaculture productivity, and significant economic losses for fish farmers. Early detection and accurate treatment are essential to prevent further disease spread. This study aims to develop an automated freshwater fish disease detection system based on digital images using the Convolutional Neural Network (CNN) method. The CNN model is designed with a layered architecture comprising convolutional, pooling, and dense layers, and is trained using the TensorFlow framework with stepwise parameter optimization. The dataset consists of 700 fish images covering six disease categories and healthy fish, all of which underwent preprocessing and data augmentation. Evaluation results show that the system achieved an accuracy of 77%, precision of 77%, recall of 76%, and F1-score of 75%, based on the confusion matrix. These results indicate that the model performs well in distinguishing between healthy and infected fish. Therefore, this system has strong potential to be applied in aquaculture environments as a fast, efficient, and technology-based tool for fish disease diagnosis.