TomaFDNet: A multiscale focused diffusion-based model for tomato disease detection.

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Tomato cultivation is vital globally, but foliar diseases threaten yield and quality.
• Current disease detection methods struggle with multi-scale features, small targets, and background complexity.

Purpose of the Study:

• To develop an advanced model for accurate multi-scale tomato leaf disease detection.
• To address limitations in existing methods for identifying small diseased areas and complex backgrounds.

Main Methods:

• Introduction of the Tomato Focus-Diffusion Network (TomaFDNet).
• Utilization of a multi-scale focus-diffusion network (MSFDNet) and an efficient parallel multi-scale convolutional module (EPMSC).
• Enhancement of multi-scale feature extraction for improved small target detection.

Main Results:

• TomaFDNet achieved a mean average precision (mAP) of 83.1% for Early_blight, Late_blight, and Leaf_Mold detection.
• Outperformed classical algorithms like Faster R-CNN and YOLO series.
• Demonstrated a statistically significant 4.2% mAP improvement over the YOLOv8 baseline (P < 0.01).

Conclusions:

• TomaFDNet provides a robust and accurate solution for precise tomato leaf disease detection.
• The model's architecture effectively handles multi-scale features and complex environmental conditions.
• This advancement supports enhanced agricultural productivity and reduced economic losses in tomato farming.