[2H5-GS-13] AI application: Image processing

[2H5-GS-13-02] Study of plant judgment system for science education using Mask R-CNN

〇Yumana Hirata¹, Masashi Kasamatsu¹, Yukikazu Murakami¹, Kazuhiro Shigeta¹ (1. National Institute of Technology, Kagawa College)

Keywords:deep learning, Mask R-CNN, Image recognition

School science requires the enhancement of familiar nature experiences, and the purpose is to nurture an attitude of protecting plants and organisms in the natural world by directly touching nature. On the other hand, it is a problem that many supervisors lack knowledge and skills of plants and organisms and feel anxious about teaching. In order to solve these problems, this study uses the Mask R-CNN, one of the object instance segmentation algorithms, to learn the plants in the image, and a plant judgment system that can be used by both students and teachers. Specifically, among the approximately 10 kinds of plants listed in the elementary school learning guidebook, discrimination was performed on the images of the flowers of Himejoon, mandarin orange, green pepper, and karatachi which are difficult to distinguish.100 images of each plant were prepared and verified as learning data 70%, verification data 20%, and test data 10%.