Automated leaf disease detection and treatment recommendation using Transfer Learning

Malpe, Sachin (2019) Automated leaf disease detection and treatment recommendation using Transfer Learning. Masters thesis, Dublin, National College of Ireland.

Preview

PDF (Master of Science) Download (1MB) | Preview

Abstract

Current automated leaf disease detection models do not generalize well on unseen data and require enormous amounts of time, computational resources and data to build. On the other hand, not much has been explored on how to provide treatment recommendations after disease detection. Deep learning (DL) methods applied using transfer learning tends to drastically reduce the amount of data, time and resources needed to build this type of model which generalizes well on new data. Hence, this study utilizes pre-trained DL architectures such as VGG16, VGG19 and SqueezeNet as feature extractors along with Extreme Gradient Boosting (XGBoost) and Support Vector Machines (SVM) as the classifier to detect 27 distinct classes of (plant, disease) combination which also includes healthy plants. Out of all the hybrid models, VGG16+SVM performed the best with the highest f1 score of 96.16%, with one of the fastest overall computational time recorded using only CPU’s. This proves the viability of this approach, which can be further researched to improve this model. Treatment recommendation is provided after detection of disease by building a rest API providing the disease and treatment details in a single JSON object.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QA Mathematics > Computer software T Technology > T Technology (General) > Information Technology > Computer software Q Science > QK Botany
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Caoimhe Ní Mhaicín
Date Deposited:	11 Oct 2019 14:46
Last Modified:	11 Oct 2019 14:46
URI:	https://norma.ncirl.ie/id/eprint/3850

Actions (login required)

View Item