Pre-trained VGG16 and EfficientNetB3 for Multiclass Classification of Orange Leaf Diseases

Identifying orange leaf disease is very important to make sure that citrus fruit farms are healthy and produce a lot of fruits. In this research, we used two different deep-learning designs called VGG19 and EfficientNetB3. We made strong models that can automatically spot problems in orange leaves without help from humans. The models were trained and checked using a carefully picked range of pictures with healthy leaves, and diseased ones caused by different illnesses like Citrus Canker, Citrus Greening and so on. The learning process needed 10 and 30 rounds to show how well the model learned as time went by. Findings show that both VGG19 and EfficientNetB3 did a good job telling the difference between healthy orange leaves and those with disease. But, our research shows that EfficientNetB3 did better than VGG19. It got the best accuracy of 97.30% at round 30. This means that the better depth and simple rules of EfficientNetB3 make it work really well for handling orange leaf disease problems. These findings show how we can use these disease-finding systems in orange groves. The good results show how helpful deep learning models, especially EfficientNetB3 are in giving tools for early disease detection that work quickly and well. The study helps us understand better about growing crops accurately and shows how important it is to pick the right deep learning designs for the best results in orange leaf disease detection.