Image-Based Wheat Fungi Diseases Identification by Deep Learning

• Published in: Plants (Basel) Vol. 10; no. 8; p. 1500
• Main Authors: Genaev, Mikhail A., Skolotneva, Ekaterina S., Gultyaeva, Elena I., Orlova, Elena A., Bechtold, Nina P., Afonnikov, Dmitry A.
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 21.07.2021; MDPI
• Subjects: Accuracy; Agricultural production; Airborne microorganisms; Algorithms; Artificial neural networks; Cellular telephones; Cereals; Control methods; Crop fields; Crop yield; Datasets; Deep learning; Developmental stages; Digital imaging; Disease control; Diseases; Electronic devices; Fungal diseases; Fungi; Fungicides; Hash based algorithms; Labeling; Labels; Leaf rust; Machine learning; Medical imaging; Neural networks; Object recognition; Pathogens; Pesticides; Plant diseases; Powdery mildew; Seedlings; septoria; Shoots; Stem rust; Stripe rust; Training; Wheat; yellow rust
• ISSN: 2223-7747; 2223-7747
• DOI: 10.3390/plants10081500

Diseases of cereals caused by pathogenic fungi can significantly reduce crop yields. Many cultures are exposed to them. The disease is difficult to control on a large scale; thus, one of the relevant approaches is the crop field monitoring, which helps to identify the disease at an early stage and take measures to prevent its spread. One of the effective control methods is disease identification based on the analysis of digital images, with the possibility of obtaining them in field conditions, using mobile devices. In this work, we propose a method for the recognition of five fungal diseases of wheat shoots (leaf rust, stem rust, yellow rust, powdery mildew, and septoria), both separately and in case of multiple diseases, with the possibility of identifying the stage of plant development. A set of 2414 images of wheat fungi diseases (WFD2020) was generated, for which expert labeling was performed by the type of disease. More than 80% of the images in the dataset correspond to single disease labels (including seedlings), more than 12% are represented by healthy plants, and 6% of the images labeled are represented by multiple diseases. In the process of creating this set, a method was applied to reduce the degeneracy of the training data based on the image hashing algorithm. The disease-recognition algorithm is based on the convolutional neural network with the EfficientNet architecture. The best accuracy (0.942) was shown by a network with a training strategy based on augmentation and transfer of image styles. The recognition method was implemented as a bot on the Telegram platform, which allows users to assess plants by lesions in the field conditions.