Lemon Leaf Disease Classification Using CNN-based Architectures with Transfer Learning

Various pest-affected and citrus diseases of lemon leaf have become very severe in the temperate weather areas of southeast Asian countries. As a result, the cultivation of lemon and other citrus fruits has been badly affected. An efficient classification of these kinds of diseases can decrease the rate of loss by choosing proper pesticides in time. In this paper, we have applied some Transfer Learning-based Deep Learning models (DenseNet-201, ResNet-50, ResNet-152V2, and Xception) for a cost-effective classification of lemon leaf diseases. We have used our image dataset that is collected from the field level. Among the models we have used, Xception achieved a very higher overall accuracy of 94.34% and outperformed the other previous works.