Convolutional Neural Networks for the Automatic Identification of Plant Diseases

Area of Science:

• Computer Science, Artificial Intelligence, Machine Learning
• Agricultural Science, Plant Pathology, Crop Science

Background:

• Deep learning, particularly Convolutional Neural Networks (CNNs), has advanced image processing capabilities.
• Automatic crop disease identification using AI is crucial for sustainable agriculture and food security.
• Numerous applications have emerged since 2016, offering potential for expert assistance and automated screening.

Purpose of the Study:

• To survey and assess the potential of Convolutional Neural Networks (CNNs) for automatic crop disease identification.
• To analyze the profiles, implementation, and performance of existing CNN-based crop disease identification studies.
• To identify key challenges and limitations in the current research landscape.

Main Methods:

• Systematic review of 19 studies utilizing CNNs for automatic crop disease identification.
• Analysis of study methodologies, network architectures, datasets, and reported performance metrics.
• Identification of common implementation aspects and challenges across the surveyed research.

Main Results:

• CNNs demonstrate significant potential for accurate crop disease identification, contributing to agricultural advancements.
• The review highlights variations in study designs, data preprocessing, and model validation.
• Key issues identified include data scarcity, generalization capabilities, and interpretability of CNN models.

Conclusions:

• CNNs are a powerful tool for developing automated crop disease identification systems.
• Guidelines are provided to enhance the operational use of CNNs in real-world agricultural settings.
• Future research should focus on improving model robustness, data efficiency, and addressing practical deployment challenges.