<![CDATA[Fish diseases represent a major challenge in the aquaculture industry as this phenomenon frequently leads to significant economic losses. Manual disease identification requires specialized expertise and is time-consuming in the field. Therefore, this study aims to implement the MobileNetV3Large Deep Learning architecture to automatically identify eight types of fish conditions. This research dataset utilizes 2,400 digital images distributed evenly across eight fish condition categories. Each class consists of 300 image samples, including Bacterial Red disease, Aeromoniasis, Bacterial gill disease, EUS Disease, Fungal diseases Saprolegniasis, Parasitic diseases, White tail disease, and a Healthy Fish group. The dataset was sourced from https://www.kaggle.com/datasets/irfanulhuda/fish-disease-detection-dataset. These conditions include bacterial, fungal, viral, and parasitic infections, as well as healthy fish conditions. The research methodology applies Transfer Learning techniques combined with Fine-Tuning optimization on the last 70 layers. The methodology applies a transfer learning strategy with a data split of 80% for training, 10% for validation, and 10% for testing. This step was taken to adapt the model's weights to the visual characteristics of the fish disease images. The process was evaluated using the Adam optimization function and the Categorical Cross-Entropy loss function. Experimental results demonstrate highly superior model performance on the test data. The MobileNetV3Large model successfully achieved a test accuracy of 92.92% with a loss value of 0.2099. Furthermore, evaluation through the Confusion Matrix and ROC curves yielded an average AUC value of 1.00 across the majority of classes. This figure indicates that the model possesses exceptionally high discrimination capacity and sensitivity. In conclusion, the computational efficiency of the MobileNetV3Large architecture makes this system a highly potential solution. Researchers can implement this model on mobile devices to assist fish farmers in diagnosing diseases quickly and accurately directly at the aquaculture sites]]>
