Improved EfficientNet for corn disease identification.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Corn diseases significantly impact crop yield and quality, necessitating accurate real-time identification.
• Complex backgrounds and variations in disease appearance challenge existing convolutional neural network models.
• Lightweight models often compromise accuracy for real-time performance in disease recognition.

Purpose of the Study:

• To develop an accurate and efficient model for corn disease recognition in real-world conditions.
• To overcome the limitations of existing models in handling complex backgrounds and variations.
• To provide a practical solution for timely disease prevention and control, enhancing crop yield and farmer income.

Main Methods:

• Proposed FCA-EfficientNet, building upon EfficientNet architecture.
• Integrated a fully-convolution-based coordinate attention module for enhanced spatial information acquisition.
• Employed an adaptive fusion module to integrate multi-scale image information and reduce background interference.

Main Results:

• FCA-EfficientNet demonstrated superior performance in accuracy, precision, recall, and F1 score compared to other deep learning models.
• The model features a low parameter count (3.44M) and Flops (339.74M), outperforming most lightweight networks.
• Achieved an average recognition speed of 92.88ms on an Android device, meeting real-time application requirements.

Conclusions:

• FCA-EfficientNet accurately identifies corn diseases in realistic environments.
• The model offers a practical balance between accuracy and real-time performance for agricultural applications.
• Contributes to timely disease management, leading to improved crop yield and economic benefits for farmers.