PLANT DISEASE DETECTION USING DEEP LEARNING

Abstract

Agriculture, also known as Farming, is the science or practice of raising crops. And the whole world is dependent on it and around 38% of the world is dependent on it. So, due to this, its productivity rate should be high. The productivity rate of a plant is affected by the disease in a plant. So that’s why plant disease should be detected at an early stage. For this, Farmers generally hire an agricultural expert who detects the disease using the naked eye and also they use instruments as well which are very expensive and which is not possible for all the farmers to afford it. There’s another way to detect a plant disease, by using Artificial Intelligence(AI). Machine Learning(ML), Deep Learning(DL) which is a sub-branch of AI, is used in agriculture in order to detect disease in a plant. So, in this work, a Dense-INC model is proposed which is based on Convolutional Neural Network(CNN) and it’s inspired by DenseNet and InceptionNet. This model is trained on the “Plant Pathology 2020: FGVC7 dataset” and “Plant Pathology 2021: FGVC8 dataset”. The proposed model is first trained with 4 optimizers: Adam, Adadelta, Adagrad, and SGD with momentum and when I compare results then shows that Adagrad gives better results than other optimizers. To further evaluate the performance of the proposed model, the proposed model is further compared with two CNN-based models with Adagrad optimizer which are already been proposed. And the results show that the proposed model gives better results than two other CNN-based models and it’s able to detect the disease with a low error rate.