Dual-branch collaborative learning network for crop disease identification.

Area of Science:

• Agricultural Science
• Computer Vision
• Deep Learning

Background:

• Crop diseases pose a significant threat to agricultural yield and global food security.
• Traditional manual crop disease monitoring lacks the efficiency and accuracy required for modern intelligent agriculture.
• Advancements in computer vision and deep learning offer potential solutions for automated crop disease detection.

Purpose of the Study:

• To develop an advanced deep learning network for accurate and efficient crop disease identification.
• To address the limitations of manual monitoring and existing automated methods in agriculture.

Main Methods:

• Proposed a dual-branch collaborative learning network (DBCLNet) for crop disease identification.
• Employed a dual-branch collaborative module with multi-scale convolutional kernels to extract global and local image features.
• Integrated a channel attention mechanism within each branch to refine feature representation.
• Utilized a feature cascade module by cascading dual-branch modules for multi-level abstract feature learning.

Main Results:

• DBCLNet achieved superior classification performance on the Plant Village dataset for 38 crop disease categories compared to state-of-the-art methods.
• Achieved high performance metrics: 99.89% Accuracy, 99.97% Precision, 99.67% Recall, and 99.79% F-score for identifying 38 crop diseases.

Conclusions:

• The proposed DBCLNet effectively utilizes global and local image features through its dual-branch architecture and attention mechanisms.
• DBCLNet demonstrates significant potential for intelligent agriculture by providing highly accurate automated crop disease identification.
• The method offers a robust solution to enhance crop monitoring, improve yield, and bolster food security.