An Analysis of Plant Diseases Identification Based on Deep Learning Methods.

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Plant diseases significantly impact crop yield and agricultural economics, necessitating early and accurate identification.
• Current deep learning models for plant disease classification often lack real-world field data and struggle with detecting multiple diseases simultaneously.
• Object detection methods offer a promising solution to overcome these limitations in practical agricultural applications.

Purpose of the Study:

• To develop and evaluate deep learning-based object detection models for identifying apple leaf diseases in real field environments.
• To address the limitations of existing datasets by creating a new annotated dataset of apple leaf diseases.
• To compare the performance of Faster R-CNN and YOLOv3 architectures for apple leaf disease detection.

Main Methods:

• Construction of an annotated apple leaf disease dataset captured in a real field environment.
• Training and evaluation of Faster R-CNN and YOLOv3 deep learning architectures on the constructed dataset.
• Comparative analysis of model performance using various evaluation metrics.

Main Results:

• Deep learning models, specifically YOLOv3 and Faster R-CNN, demonstrate feasibility for detecting apple leaf diseases in field conditions.
• The study highlights the strengths and weaknesses of each tested deep learning architecture.
• The newly constructed dataset aids in compensating for the lack of real-world data in existing plant disease research.

Conclusions:

• Object detection methods based on deep learning, such as YOLOv3 and Faster R-CNN, are effective for identifying multiple apple leaf diseases in complex field settings.
• The developed dataset and evaluated models contribute to advancing automated crop disease management.
• Further research can build upon these findings to enhance the practical application of AI in agriculture for disease surveillance.