<abstract xmlns="http://www.w3.org/1999/xhtml">

<p>Modern reviews of challenges related to deep learning application in agriculture mention restricted access to open datasets with high-resolution natural images taken in field conditions. Therefore, artificial intelligence solutions trained on these datasets containing low-resolution images and disease symptoms in the advanced stage are not suitable for early detection of plant diseases. The study aims to train a convolutional neural network for apple scab detection in an early stage of disease development. In this study a dataset was collected and used to develop a convolutional neural network based on the sliding-window method. The convolutional neural network was trained using the transfer-learning approach and MobileNetV2 architecture tuned on for embedded devices. The quality analysis in laboratory conditions showed the following accuracy results: <italic>F</italic><sub>1</sub> score 0.96 and Cohen’s kappa 0.94; and the occlusion maps — correct classification features.</p>
</abstract>

Modern reviews of challenges related to deep learning application in agriculture mention restricted access to open datasets with high-resolution natural images taken in field conditions. Therefore, artificial intelligence solutions trained on these datasets containing low-resolution images and disease symptoms in the advanced stage are not suitable for early detection of plant diseases. The study aims to train a convolutional neural network for apple scab detection in an early stage of disease development. In this study a dataset was collected and used to develop a convolutional neural network based on the sliding-window method. The convolutional neural network was trained using the transfer-learning approach and MobileNetV2 architecture tuned on for embedded devices. The quality analysis in laboratory conditions showed the following accuracy results: F1 score 0.96 and Cohen’s kappa 0.94; and the occlusion maps — correct classification features.

Apple Scab Detection in the Early Stage of Disease Using a Convolutional Neural Network

Proceedings of the Latvian Academy of Sciences. Section B. Natural, Exact, and Applied Sciences.

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Modern reviews of challenges related to deep learning application in agriculture mention restricted access to open datasets with high-resolution natural images...

Apple Scab Detection in the Early Stage of Disease Using a Convolutional Neural Network

Online veröffentlicht: 14. Okt. 2022

Seitenbereich: 482 - 487

Eingereicht: 26. Aug. 2021

Akzeptiert: 15. Juli 2022

DOI: https://doi.org/10.2478/prolas-2022-0074

Schlüsselwörter
agricultural informatics, deep learning, image recognition, precise horticulture,

© 2022 Sergejs Kodors et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Apple Scab Detection in the Early Stage of Disease Using a Convolutional Neural Network

Online veröffentlicht: 14. Okt. 2022

Seitenbereich: 482 - 487

Eingereicht: 26. Aug. 2021

Akzeptiert: 15. Juli 2022

DOI: https://doi.org/10.2478/prolas-2022-0074

Schlüsselwörteragricultural informatics, deep learning, image recognition, precise horticulture,

© 2022 Sergejs Kodors et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Schlüsselwörter
agricultural informatics, deep learning, image recognition, precise horticulture,