Transfer Learning Approach for Plant Leaf Disease Detection Using CNN with Pre Trained Feature Extraction Method Mobilnetv2

Abstract

Economy of Bangladesh mostly depended on agriculture. As a small country our population is over than the limit. We are a developing country. For national GDP growth must be to increase the production. Every year a huge amount of agricultural production losses for the crop disease. As we know, most of the farmers in our country are illiterate, have no proper knowledge about the disease, that’s why they cannot manually detect the disease. I think if you properly detect the disease at the early stage we can solve the issue. We develop a model that can classify leaf disease. We focus on 5 major production crops in Bangladesh. Using computer vision technique our farmer will get the benefit. We use convolutional neural networks for classifying images. An algorithm cannot properly capture the features of existing data that’s why we use pre-trained feature extraction method that is MobileNetv2. MobileNetv2 is very useful for mobile devices. The research contains the proportions of validation accuracy of 90.38%. This approach resulted in the agriculture sector that will help a farmer to classify disease from harvest. The main goal of our model is minimize the damage of suffering plants that can help to growth of production. By solving this issue themselves farmers can also reduce cost. Our goal is that they can cure their crop at the right time. To achieve this goal we tend to think that we should develop a way to detect the leaf disease. We collect several kinds of leaf. And then any leave can be tasted by our model. By using our model we try to reduce the leaf disease