PM-YOLO: A Powdery Mildew Automatic Grading Detection Model for Rubber Tree.

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Powdery mildew significantly impacts rubber tree yield and quality.
• Early detection is crucial but conventional methods are inefficient.
• Automated disease detection is needed for timely intervention.

Purpose of the Study:

• To develop a deep learning model for accurate and efficient detection of powdery mildew in rubber trees.
• To create a comprehensive dataset for training the detection model.
• To implement an automatic grading algorithm for assessing disease severity.

Main Methods:

• Constructed a dataset of 6200 rubber tree images with 38,000 annotations.
• Developed PM-YOLO, a deep learning model based on the YOLO framework.
• Integrated a Feature Focus and Diffusion Mechanism (FFDM) and Dimension-Aware Selective Integration (DASI) module.
• Proposed an automatic grading algorithm for disease severity.

Main Results:

• PM-YOLO achieved 86.9% mean average precision (mAP) and 85.6% recall.
• Outperformed standard YOLOv10 by 7.6% mAP and 8.2% recall.
• Demonstrated accurate detection in complex backgrounds and effective grading.

Conclusions:

• The proposed PM-YOLO model provides accurate, real-time detection of rubber tree powdery mildew.
• The automated grading system offers an effective solution for early diagnosis and management.
• This deep learning approach addresses the limitations of traditional disease detection methods.