Detecting Pests on Tomato Plants using Convolutional Neural Networks

Vatti, Srivenkateswara Rao (2020) Detecting Pests on Tomato Plants using Convolutional Neural Networks. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (4MB) | Preview
Preview	PDF (Configuration manual) Download (980kB) | Preview

Abstract

Image processing is widely used in various industries around the globe such as bioinformatics, space, weather forecasting, disease diagnosis and so on. With the recent advancements in the field of deep learning and artificial intelligence (AI), GPU-powered deep learning frameworks such as TensorFlow, Pytorch, Microsoft Cognitive Toolkit and others, many challenging problems in computer vision such as image classification, object detection and many more can be solved. The current research emphasises on detecting and classifying the pest that is formed on tomato plants. The analysis carried out in this paper is based on convolutional neural networks(CNN). The usage of scouting robots in the agriculture industry is increasing than ever before and many big organizations are investing large amounts of money on the same. The primary goal of developing robotic solutions through AI is to reduce the manpower that is utilized for harvesting purposes. As part of the current research, initial blocks from a set of CNN models such as VGG16, VGG19, Xception, ResNet50 and Inception V3 along with additional convolutional layers are applied on the dataset chosen and results are evaluated with the help of various standard metrics. Maximum classification accuracy of 0.95 is achieved through a CNN model, which is implemented with a set of convolutional layers from the VGG16 and additional layers which are added explicitly. Also, a comparative analysis is carried out with other models developed with transfer learning. Detection and classification of pests or insects for the selected dataset is a challenging task since the size of them is very minute. The results that are obtained in terms of accuracy for detecting and classifying the type of pest/insect can be crucial in integrating the developed models to a scouting robot which can accept an image as an input and identifies the type of pest in that image. Early detection of pest/insects can minimise the usage of pesticides and increase the overall productivity of the crop. The integration of the models implemented with scouting robots is not covered as part of this research.

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 S Agriculture > S Agriculture (General)
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Dan English
Date Deposited:	25 Jan 2021 16:04
Last Modified:	25 Jan 2021 16:04
URI:	https://norma.ncirl.ie/id/eprint/4478

Actions (login required)

View Item