## Early diagnosis of shrimp diseases with AI Shrimp production is an economically important sector, but it is often threatened by disease outbreaks. Timely and accurate diagnosis is crucial to mitigate losses. A recent study has proposed a deep learning-based approach for the automated classification of shrimp diseases. The system uses a dataset of 1,149 images divided into four disease classes. ## Implementation details Six pre-trained deep learning models were used: ResNet50, EfficientNet, DenseNet201, MobileNet, ConvNeXt-Tiny, and Xception. The images were pre-processed with background removal techniques and standardized using the Keras pipeline. To improve model robustness, the Fast Gradient Sign Method (FGSM) was used for adversarial training. Data augmentation strategies such as CutMix and MixUp were implemented to reduce overfitting and improve generalization. ## Results and interpretability The ConvNeXt-Tiny model achieved the highest accuracy, scoring 96.88% on the test dataset. To support interpretability, post-hoc explanation methods such as Grad-CAM, Grad-CAM++, and XGrad-CAM were applied to visualize the model's attention regions. These promising results pave the way for more efficient and automated monitoring systems in the aquaculture sector, contributing to more sustainable shrimp production.