Betel Leaf Disease Recognition Using Deep Learning

Abstract

In South Asia, a large number of people consume betel life every day and it becomes a part of the daily life of this reason for many years. Betel leaf is also used as a remedy and, it's a great source of nutrition. In the international market demands for betel leaf is increasing incredibly. Plants of betel leaf are highly sensitive, so its cultivation is very laborious work. Like other South Asian countries, farmers of Bangladesh also face a great loss for the diseases of betel leaf every year. Diseases of betel plants decrease leaves quality. Recognition of betel leaf diseases became a crucial task to meet the rising demand in local and international markets and increase the quality of leaf. We will recognize the diseases of betel leaf using deep learning. First, we study some online articles, journals, and, related papers then we talk to local farmers and visit their lands; we find two common diseases of betel leaf which are foot rot and, leaf rot. Then we collect images of healthy and affected betel leaves. We also collected some images of unknown diseases. After data collecting images, we resized images and create a dataset of four classes. We applied deep learning models in our dataset to recognize the disease of betel leaf. Now deep learning is used in various recognition related systems. We have proposed a CNN model named BLCNN, which achieved 90.75% training accuracy and 89.44% test accuracy. Besides BLCNN we have also modified 3 pre-trained models for this study. Among these, EfficientNet B0 has performed better than other models. It has achieved 96.77% accuracy in classifying new images.