Enhancing Orange Crop Health: A Hybrid Deep Learning Approach for Precise Multi-Classification of Leaf Diseases

This research paper introduces a novel hybrid deep learning model that combines the strengths of EfficientNet B3 and VGG16 to address the challenge of accurately classifying orange leaf diseases into three distinct categories: healthy, sooty mold sick, and citrus canker sick. The modern farming industry is clouded by higher risks of plant disease infestation that can both greatly reduce yields and impair quality. This has given rise to the growing need for advanced technologies that could quickly and precisely detect any outbreak of disease. Utilizing the refined functionalities of convolutional neural networks (CNNs), our model, in particular, will make the process of disease classification automated and with this level of automation both the speed and accuracy of crop management practices will be enhanced. The methodology is based on the taking of a large dataset of orange leaf images and then data preprocessing, training, and evaluation of possibly a hybrid model. We gathered evidence that the proposed model provides a 97.87% overall accuracy level, which is significantly better than the individual rates of EfficientNet B3 and VGG16. The EfficientNet B3 is considered to be a highly efficient model; and, the VGG16 is popular for its robust characteristic in the extraction of features. The hybrid model fitted the bills by providing high accuracy, recall, and F1 scores across the disease groups thereby followed to augment the diagnosis rates to a great extent.