Enhanced multiscale plant disease detection with the PYOLO model innovations

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Accurate plant disease detection is vital for agricultural safety, product quality, and environmental protection.
• Existing methods face challenges due to diverse disease scenarios and complex backgrounds.
• Need for robust and efficient plant disease detection models.

Purpose of the Study:

• To develop an advanced plant disease detection model named PYOLO.
• To enhance feature fusion, scale-specific feature extraction, and background/target perception capabilities.
• To improve the overall accuracy and robustness of plant disease detection.

Main Methods:

• Optimized PAN structure with weighted bidirectional feature pyramid network (BiFPN) for enhanced feature fusion.
• Redesigned EC2f structure with dynamic kernel size adjustment for multi-scale feature capture.
• MHC2f mechanism with self-attention for improved perception of complex backgrounds and targets.

Main Results:

• The PYOLO model demonstrated superior performance in plant disease detection.
• Achieved a 4.1% increase in mean Average Precision (mAP) compared to YOLOv8n.
• Validated effectiveness in handling diverse and complex detection scenarios.

Conclusions:

• PYOLO offers a significant advancement in automated plant disease detection.
• The model's architectural improvements lead to enhanced accuracy and robustness.
• PYOLO provides a promising solution for safeguarding agriculture.