search for




 

Detection of Scirtothrips with deep neural networks
Journal of the Korean Data & Information Science Society 2018;29:1287-97
Published online September 30, 2018
© 2018 Korean Data and Information Science Society.

Donghwan Lee1 · Kyungha Seok2

12Department of Statistics, Inje University
Correspondence to: Professor, Institute of Statistical Information, Department of Statistics, Inje University, Gyungnam 50834, Korea. E-mail: statskh@inje.ac.kr
Received August 9, 2018; Revised December 17, 2018; Accepted September 18, 2018.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
In this paper, we study on a detection of Scirtothrips dorsalis Hood, which is classified as a major insect in citrus farming. The detection is based on the deep neural networks, specifically the Faster R-CNN (faster regions with CNN) model based on CNN (convolutional neural network), with the yellow sticky trap image data (250×150mm, 5472×3648pixels). It was found that the model performance becomes unstable when the object is too small and rare. In order to solve this problem, we use pretrained weights to set the initial value of the model, as well as we select hyperparameters by grid search. Result shows that our proposed model has an high AUC (area under curve) value 0.91. We expect that it would be possible to know more precisely the lifespan of the Scirtothrips dorsalis Hood and to control them more precisely through our proposed model.
Keywords : Convolutinal network, deep learning, Faster R-CNN, object detection, Scirtothrips dorsalis Hood.