檜垣, 泰彦

須貝, 康雄

HIGAKI, Yasuhiko

SUGAI, Yasuo

ディープラーニング

自動マスキング

セグメンテーション

U-Net

Plant specimen image

Deep learning

Automatic masking

Segmentation

U-Net

[Abstract] the image processing required to digitize images of plant specimens. Digitized images of plant specimens are very diverse and contain visual noise and undisclosed information. As a pre-processing step for downstream deep learning applications, systematic removal of noise and segmentation of plant bodies is required. In this study, we developed a workflow that uses deep learning to segment and remove background from plant bodies in plant specimen images. A combination of automatic and manual tools was used to generate ground truth masks, train U-Net, and automatically segment the specimen images with good results.