Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/412306
Title: | Machine Learning Approaches for Crop Leaf Disease Detection and Classification |
Researcher: | Darak Mahesh Shyamsunder |
Guide(s): | Deshmukh Nilesh K. |
Keywords: | Computer Science Computer Science Theory and Methods Engineering and Technology |
University: | Swami Ramanand Teerth Marathwada University |
Completed Date: | 2022 |
Abstract: | In India s economy, agriculture plays an important role. Agriculture provides a living newlinefor over 70% of the population in India. One of the greatest issues confronting the newlineagricultural industry is the requirement for accurate and timely identification of newlinecrop-damaging diseases. Diseases have an impact on crop quality and have the potential to newlinewipe out hectares of agricultural production, resulting in significant losses for farmers. newlineCurrent diagnostic approaches are time intensive and need the presence of highly qualified newlineexperts to study the damaged crop, comprehend the symptoms, diagnose the disease, and newlinesuggest appropriate treatments. The limits of such techniques have compelled researchers newlineto seek for alternate methods for detecting and classifying diseases at an early stage. newlineSmart farming combined with appropriate infrastructure can aid in addressing and newlineresolving such problems. newlineMachine Learning has showed significant promise in recent years in detecting and newlineclassifying patterns in related fields of study. The current study seeks to compare the newlineperformance of Convolutional Neural Network (CNN) methods and architectures such as newlineInceptionv3, VGG16, and RasNet50 with data augmentation and transfer learning to newlinetraditional machine learning techniques such as Support Vector Machine (SVM), random newlineforest, and analyse their usefulness in recognising and classifying Cotton leaf diseases in newlineterms of accuracy, precision, recall, and training time. The models were trained using a newlinemanually obtained database from a farm and a government agency, which included four newlineunique classes of images, including healthy plant. The Inceptionv3 architecture of CNN newlinewith transfer learning performed the best of the models created, obtaining an overall newlineaccuracy of 94 percent and thus satisfying the need for a more effective and robust newlineclassification model. In addition, the performance of the created models was observed to newlineimprove as the amount of training data increased. newlineThe results achieved via transfer learning algorithms on CNN architecture |
Pagination: | 85p |
URI: | http://hdl.handle.net/10603/412306 |
Appears in Departments: | Department of Computer Science |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 50.35 kB | Adobe PDF | View/Open |
02_certificate.pdf | 42.83 kB | Adobe PDF | View/Open | |
03_abstract.pdf | 63.61 kB | Adobe PDF | View/Open | |
04_decleration.pdf | 42.41 kB | Adobe PDF | View/Open | |
05_acknowledgement.pdf | 43.41 kB | Adobe PDF | View/Open | |
06_contents.pdf | 44.18 kB | Adobe PDF | View/Open | |
07_list_of_tables.pdf | 42.24 kB | Adobe PDF | View/Open | |
08_list_of_figures.pdf | 43.31 kB | Adobe PDF | View/Open | |
09_abbrevations.pdf | 42.47 kB | Adobe PDF | View/Open | |
10_chapter 1.pdf | 711.72 kB | Adobe PDF | View/Open | |
11_chapter 2.pdf | 81.9 kB | Adobe PDF | View/Open | |
12_chapter 3.pdf | 1.59 MB | Adobe PDF | View/Open | |
13_chapter 4.pdf | 242.07 kB | Adobe PDF | View/Open | |
14_conclusion.pdf | 45.5 kB | Adobe PDF | View/Open | |
15_bibliography.pdf | 55.86 kB | Adobe PDF | View/Open | |
80_recommendation.pdf | 55.13 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: