Advancing Citrus Disease Diagnosis: Application of EfficientNetB3 for Precise Classification of Orange Tree Pathologies

This study investigates the application of the EfficientNetB3 convolutional neural network model for the classification of four major diseases affecting orange trees: canker, greasing, scab, and greening. since these citrus diseases are highly detrimental to the citrus output, and producing revenues, therefore, diagnosis of these diseases must be taken as a top priority and should be done as quickly as possible for effective management approaches to be implemented. By utilizing a dataset of leaf images that is quite comprehensive, the research relies on the EfficientNetB3 model, which is a classifier with different parameters, to come up with an effective classification system for the featured diseases. The research methodology consists of the data collection process, preprocessing, model training, and validation by using a model dataset split with a split of the dataset into training, validation, and test sets, to measure the robustness of the model. The evaluation of the EfficientNetB3 model is done based on accuracy, precision, recall, and Fl-score metrics, and based on the obtained results the model proved to be very effective and had an overall accuracy of 92.25% in disease classification. This proves how great deep learning models like EfficientNetB3 can be beneficial to the agricultural industry for outdoor plant disease diagnostics at high speed, even without practical work touching the thing or missing an essential quality. While the study concerns the utilization of artificial intelligence in plant pathology, it also reveals a certain level of the model’s superiority because it can improve plant disease management programs. Along with these findings, the research also identifies shortcomings and barriers. Among these are nil observations which might be misguided by the symptoms of similar diseases, therefore, indicating areas where model refinements and possibly application can be done.