AUTHOR=Yang Guofeng , He Yong , Yang Yong , Xu Beibei 

TITLE=Fine-Grained Image Classification for Crop Disease Based on Attention Mechanism

JOURNAL=Frontiers in Plant Science

VOLUME=Volume 11 - 2020

YEAR=2020

URL=https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2020.600854

DOI=10.3389/fpls.2020.600854

ISSN=1664-462X

ABSTRACT=Fine-grained image classification is challenging because it is difficult to find discriminant features. It is not easy to find the subtle features that fully represents the object. Crop disease fine-grained classification faces visual disturbances such as light, fog, overlap, and jitter. To explore the influence of features of crop leaf images on the classification results, the classification model pays more attention to some more discriminative regions of the image, and at the same time improves the classification accuracy of the model in complex scenes. This paper proposes a novel attention mechanism, which can effectively utilize the informative region of the image, and use transfer learning to quickly construct several fine-grained image classification models of crop diseases based on attention mechanism. This paper uses 58,200 crop leaf images as a dataset, including 14 different crops and 37 different categories of healthy/diseased crops. Among them, there is a strong similarity to different diseases of the same crop. The NASNetLarge fine-grained classification model based on the attention mechanism achieved the best classification effect, and the F1 value reached 93.05%. The results show that the attention mechanism can effectively improve the fine-grained classification of crop disease images.