Efficient attention-based CNN network (EANet) for multi-class maize crop disease classification.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Maize leaf diseases significantly impact crop yield and quality.
• Accurate disease diagnosis is challenging due to complex field conditions and background noise.
• Existing automated methods often struggle with realistic environmental variations.

Purpose of the Study:

• To develop an automated system for accurate multi-class maize crop disease identification.
• To enhance feature representation and disease localization in challenging environments.
• To overcome limitations of existing automated disease detection methods.

Main Methods:

• An end-to-end Convolutional Neural Network (CNN) architecture, Efficient Attention Network (EANet), was developed.
• A spatial-channel attention mechanism was integrated to focus on disease-affected areas.
• The model was trained using focal loss for class imbalance and transfer learning for generalization.

Main Results:

• The EANet model achieved an overall accuracy of 99.89% in categorizing maize crop diseases.
• The attention mechanism effectively highlighted disease-relevant information while mitigating background noise.
• The model demonstrated superior performance compared to conventional CNNs under varied environmental conditions.

Conclusions:

• The proposed EANet model offers a highly accurate and robust solution for maize leaf disease identification.
• The integration of attention mechanisms significantly improves disease detection in complex field settings.
• This approach can aid in timely crop monitoring and management to preserve yield and quality.