Identification of Plant Leaves Diseases Using Machine Learning

 




 

Khoo, Vincent Khai Jie (2020) Identification of Plant Leaves Diseases Using Machine Learning. Final Year Project (Bachelor), Tunku Abdul Rahman University College.

[img] Text
VincentKhooKhaiJie_Full Text.pdf
Restricted to Registered users only

Download (4MB)

Abstract

Plant is an importance species on Earth which able to provide a lot of benefit to human beings and animals. As plant is also a form of organisms which will easy be infected by diseases. Hence, with the diseases occur on plant especially crops, it will cause the growth of economic of a country and it also will affect the incomes of the farmer. Therefore, plant diseases are an importance factor that has been drawn attention by scientist and expert of plant pathology in all around of world. Since plant diseases is hard to be identify with naked eyes, consequently, farmers need to consult with expertise and scientist which is costly and time consuming and the results that investigate by them may be incorrect. Moreover, in the prior works, there is extra implementation of manual pre-processing technique to the images dataset and causes long processing time for the whole process of identification of the plant diseases. The objective of this research project is to develop a plant leaves diseases identification system on machine learning. This project is a modern plant diseases identification tool which need to apply machine learning algorithm to it. The images dataset of the plant diseases along with image annotation information will prepare and Mask R-CNN algorithm was implemented as the machine learning system in this project. In the Mask R-CNN, Resnet 50 has chosen as the backbone of the model which provide deep learning in the process. Then, features maps will be produced and Region Proposal Network (RPN) will localise the diseases area with anchor boxes. After that, ROI Align will used to identify the Region of Interest of target diseases area with rescale and localise the anchor boxes correctly to the corresponding disease area. Moreover, the bounding box and class of diseases will be form in the Fully Connected Layer and the pixel-wise mask will formed in the Fully Convolutional Network Layer to specific the position of the diseases area in the image. Finally, the trained system will be identified with plant diseases testing set for verify its mean Average Precision to identify and classify different types of diseases. This project was developed in Kaggle with using Python as the programming language. Kaggle is a public or open source online compiler that allow user to create their own machine learning system. In this project 10 types of diseases with overall of 1,000 images will be used as the image dataset. In this project, the mean Average Precision has achieved 0.7169 by implementing Mask RCNN to identify different types of diseases. In this research project, using Mask R-CNN as training method in identify the plant diseases could help farmer to perform early prediction of the diseases, identify the plant diseases correctly and also reduce the consulting hours with the experts which is expensive and time wasting.

Item Type: Final Year Project
Subjects: Technology > Mechanical engineering and machinery
Technology > Electrical engineering. Electronics engineering
Faculties: Faculty of Engineering and Technology > Bachelor of Mechatronics Engineering with Honours
Depositing User: Library Staff
Date Deposited: 24 Apr 2020 16:24
Last Modified: 24 Apr 2020 16:24
URI: https://eprints.tarc.edu.my/id/eprint/14322