Implementing ResNet for Early Detection and Classification of Fish Skin Diseases

Le résumé fourni par la source

Background: Fish diseases pose significant challenges in aquaculture, impacting health and productivity. Timely detection is important for effective management. This study explores the application of a ResNet-20 deep-learning model for classifying various fish skin diseases. Methods: We utilized a dataset sourced from Kaggle, comprising 442 images categorized into seven groups: healthy fish and six disease types. To improve variety, images were scaled up to 224 ´ 224 pixels. Training (70%), validation (15%) and testing (15%) sets make up the dataset partition. Result: The overall accuracy of the model was 82.35%. Strong performance was shown by classification metrics, especially for healthy fish and Aeromoniasis. For the majority of disease categories, AUC values above 0.9 were found using ROC curves, indicating effective classification. The ResNet-20 model effectively detects fish diseases, showcasing the potential of deep learning in aquaculture applications. This research provides insights into the strengths and limitations of the model. Future work should focus on expanding the dataset and exploring additional neural network architectures to enhance accuracy and generalization.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.