Classification of Grape Leaf Diseases Using Convolutional Neural Network

Chong, Wai Jun (2023) Classification of Grape Leaf Diseases Using Convolutional Neural Network. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Full Text _CHONG WAI JUN.pdf Restricted to Registered users only Download (10MB)

Abstract

Productivity of grape plants are negatively affected by diseases, while visual inspection by trained farmers are expensive and time consuming, the judgement provided by each individual at different expert level is subjective. Previous works on plant diseases classification demonstrated that Convolutional neural network (CNN) is able to outperform other machine learning methods in classification accuracy. In this project, a new CNN algorithm which is inspired by InceptionV3 model, has been proposed to identify 4 common grape leaf diseases. Dataset from PlantVillage were used for model training. The dataset are then augmented to 12270 number of images, and then divided into three datasets, which are training, validation and testing. Image preprocessing of image resizing and formatting, and smooth filter operation are done before the data is fed into the CNN model for feature classification. The highest classification accuracy of the CNN model developed is 99.949% on testing dataset. Then, performance of proposed has been evaluated using standard performance evaluation metrics. Experimental results have shown that the model is able to achieve good accuracy rate for grape leaf diseases classification problem.