Explainable deep learning reproduce a "professional eye" on the diagnosis of internal disorders in persimmon fruit

Takashi Akagi, Masanori Onishi, Kanae Masuda, Ryohei Kuroki, Kohei Baba, Kouki Takeshita, Tetsuya Suzuki, Takeshi Niikawa, Seiichi Uchida, Takeshi Ise

Research output: Contribution to journalArticlepeer-review

Abstract

Recent rapid progress in deep neural network techniques have allowed various object recognitions and classifications, which often exceed the performance with human eyes. In plant biology and crop sciences, some deep neural network frameworks have been applied mainly to effective and rapid phenotyping. In this study, beyond simple optimizations of phenotyping, we propose an application of deep neural networks to make image-based internal disorder diagnosis that is hard even for experts, and to visualize the reasons of each diagnosis to provide biological interpretations. Here we exemplified classification of calyx-end cracking in persimmon fruit by using five convolutional neural network models with various layer structures, and examined potential analytic options involved in the diagnostic qualities. With 3,173 visible RGB images from the fruit apex side, the neural networks successfully made the binary classification of each degree of disorders, with up to 90% accuracy. Furthermore, feature visualizations, such as Grad-CAM and LRP, visualize the image regions that contribute to the diagnosis. They suggest that specific patterns of color unevenness, such as in fruit peripheral area, can be indexes of calyx-end cracking. These results not only provided novel insights into indexes of fruit internal disorders but proposed potential applicability of deep neural networks in plant biology.

Original languageEnglish
JournalPlant & cell physiology
DOIs
Publication statusE-pub ahead of print - Aug 26 2020

Fingerprint Dive into the research topics of 'Explainable deep learning reproduce a "professional eye" on the diagnosis of internal disorders in persimmon fruit'. Together they form a unique fingerprint.

Cite this