Deep Learning-based Apple Leaf Disease Detection using ResNet34: A Robust Approach for Agricultural Diagnosis

Apple leaf diseases are one of the most challenging factors to agriculture production and farming as a whole as they result in huge losses and affect the economy. This study presents a deep learning approach using the ResNet34 architecture to detect and classify apple leaf diseases into three categories: Apple Black Rot, Apple Cedar Rust, and Apple Scab. The dataset with 480 high-resolution images was pre-processed through reshaping, normalization and data augmentation in order to improve the model performance. ResNet34 was fine-tuned using transfer learning for the purpose of feature extraction and classification. The model was validated to have an overall accuracy of 86% in disease detection, showing the model’s effectiveness in disease diagnosis. The three-fold cross-validation was used to generate the performance evaluation metrics such as precision, recall, F1-score, and the confusion matrix in order to validate the model’s outcome for the three classes of diseases. To prevent overfitting, dropout, batch normalization, and early stopping were employed, and the training procedure was made stable. The results obtained show that ResNet34 can be considered as a promising tool for the development of a system for automation of apple leaf diseases identification. There can be future work which can include the enhancement of the dataset, integrating the real time deployment systems and exploring the complex architectures to enhance the accuracy and flexibility in the real-world scenario.