An Efficient Deep Transfer Learning based Apple Leaf Disease Classification

The accurate identification and classification of plant diseases are vital for protecting the world’s food supply and the financial interests of stakeholders. While deep learning-based systems have been introduced for other crops, there is a lack of research on automating the categorization of apple leaf diseases. This study suggests utilizing transfer learning to diagnose apple leaf ailments, where a proposed model architecture is used to extract essential features for precise predictions. To address class imbalance, the system incorporates runtime data augmentation. In a comprehensive study, researchers thoroughly investigated the impact of various hyperparameters like input resolution, learning rate, and number of epochs on the outcomes. They conducted tests on the apple leaf disease subset of the ’PlantVillage’ dataset, making use of a recommended pipeline. This pipeline surpassed previous approaches and achieved an impressive accuracy of 96.21%.