Publication:
Comparison of Convolutional Neural Network Performance Against Deep Learning Architectecture With Diseased Paddy Dataset

Loading...
Thumbnail Image
Date
2020-09
Authors
Ain Nadia Alin binti Mohd Azeman
Journal Title
Journal ISSN
Volume Title
Publisher
Research Projects
Organizational Units
Journal Issue
Abstract
Paddy farmers experience a big loss in terms of time, energy, and cost when they experienced food spoilage. During harvesting, they discovered that the paddy they planted turns out to be defective due to diseases. Their wasted time energy and money could have been prevented if they could foresee these paddy diseases. Hence, Artificial Intelligence is introduced to help them reduce their losses. There are three types of paddy diseases detected in this project, they are the Bacterial Leaf Bight, Brown Spot Disease, and the Leaf Blast Disease. The aim was to train an AI model that could produce a minimum of 20% mean average precision (mAP) value and to create a parameter that can be applied to the datasets to obtain consistent results. There were four models trained and observed. The model that obtained the highest mAP value of 0.2222 was the faster RCNN Mobilenet v2, however, it also took the longest to complete the training. There were three datasets used to test the parameter that could be applied to obtain consistent results. It was observed that dataset B (contains only augmented images) achieved the highest mAP value of 0.2061 or roughly 21% when trained using the SSD Mobilenet v1 model.
Description
Interim Semester 2020/2021
Keywords
Neural Network
Citation