Tea leaf Disease Identification Using Machine learning Approach

Abstract

Agriculture remains the most important sector of Bangladesh economy, contributing 19.6 percent to the national GDP and providing employment for 63 percent of the population. Agriculture is the largest livelihood provider in Bangladesh. Most of all, for the vast rural population. Tea is one of the important crops in Bangladesh. It contributes 1 percent of total countries GDP. And it contributes 3 percent of global tea production. The tea industry employs one and a half lakh workers. Bangladesh tea industry now crossing BDT 4000 crore. But every year the tea industry faces a large amount of losses due to leaf diseases like bacterial, gray blight or fungal infections. For those diseases and fungal infection tea production decreases. Identifying diseases from tea leaf is critical and challenging work. But we have to do that to sustain tea demand worldwide. And it is also important for Bangladesh economy. Now tea worker finds those diseases by his eyes. Most of the time affected tea remains in the garden and it spread diseases tree to tree. It decreases tea production and tea quality. We need a solution for that. So, our purpose of this project is to develop a system that can detect those tea leaf diseases. We will do that by using machine learning techniques like image processing. In image processing there are some models like SVM, and CNN (convolutional neural network) based models are usually used for research. We chose SVM, CNN and VGG-16(CNN-based model) to identify tea leaf disease. Because we previously see those pre trained models work find in this kind of image. We train those models of our 75 percent of data, and we keep 25 percent data for our testing purpose. After finishing training when we tested our models, we got 98 percent accuracy