Data-Driven Solutions in Agriculture: Machine Learning for Apple Disease Classification

The agricultural sector faces significant challenges due to plant diseases, with apple orchards being particularly susceptible to various pathogens. Accurate and timely disease detection is crucial for mitigating crop loss and ensuring food security. This paper proposes an approach to apple disease detection by leveraging both - fine-tuning and feature extraction methods using advanced pre-trained Convolutional Neural Network (CNN) architectures. Specifically, we employ ResNet-50, a widely adopted deep learning model, fine-tuning it to optimize for apple disease classification. Additionally, we utilize ConvNeXt-Xlarge for feature extraction, integrating its outputs with shallow classifiers, such as Random Forest(RF), K-Nearest Neighbors (KNN), and Multilayer Perceptron (MLP), to optimize classification accuracy. Comparative evaluations were performed on the Plant Pathology dataset, featuring six categories of apple leaf disease. Results demonstrate that ConvNeXt-Xlarge coupled with MLP achieved a test accuracy of 87.50%, outperforming ResNet-50. Achieving a high degree of precision (88%) in identifying and classifying apple diseases enables early intervention and reduces manual inspection processes. The research findings highlight the potential of deep-learning strategies in transforming plant disease management and enhancing agriculture productivity.