TomatoGuard-YOLO: a novel efficient tomato disease detection method.

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Tomato crops face significant disease threats, impacting yield and global food security.
• Existing disease detection methods lack accuracy, efficiency, and scalability.

Purpose of the Study:

• To develop an advanced, lightweight, and efficient tomato disease detection framework.
• To improve upon existing YOLO architectures for enhanced performance.

Main Methods:

• Proposed TomatoGuard-YOLO framework based on an improved YOLOv10 architecture.
• Introduced Multi-Path Inverted Residual Unit (MPIRU) for feature extraction and fusion.
• Implemented Dynamic Focusing Attention Framework (DFAF) for adaptive region focus.
• Utilized Focal-EIoU loss function for improved bounding box accuracy and class balance.

Main Results:

• Achieved a mean Average Precision (mAP50) of 94.23% on a dedicated dataset.
• Demonstrated a high inference speed of 129.64 Frames Per Second (FPS).
• Resulted in an ultra-compact model size of 2.65 MB.

Conclusions:

• TomatoGuard-YOLO presents a transformative solution for intelligent plant disease management.
• The framework offers significant advancements in detection accuracy, speed, and model efficiency.
• This technology supports sustainable agriculture and enhances global food security.