BED-YOLO: An Enhanced YOLOv10n-Based Tomato Leaf Disease Detection Algorithm.

Area of Science:

• Agricultural Science
• Computer Vision
• Plant Pathology

Background:

• Tomato crops are vital economically but vulnerable to diseases, causing significant yield and quality losses.
• Traditional disease diagnosis relies on manual inspection, which is time-consuming and subjective.
• Object detection algorithms offer efficient and accurate solutions for automated crop disease identification.

Purpose of the Study:

• To develop an improved tomato leaf disease detection method using an enhanced YOLOv10n algorithm.
• To increase the accuracy and robustness of detecting common tomato diseases in diverse conditions.

Main Methods:

• A novel algorithm, BED-YOLO, was developed by modifying the YOLOv10n architecture.
• Incorporated Deformable Convolutional Network (DCN) for better handling of occlusions and irregular lesion edges.
• Integrated Bidirectional Feature Pyramid Network (BiFPN) for optimized feature fusion and small object detection.
• Added Efficient Multi-Scale Attention (EMA) mechanism to focus on disease features and reduce noise.

Main Results:

• The BED-YOLO model demonstrated improved performance over the original YOLOv10n.
• Precision increased from 85.1% to 87.2%.
• Recall improved from 86.3% to 89.1%.
• Mean Average Precision (mAP) rose from 87.4% to 91.3%.
• The model showed strong practical applicability in natural field conditions.

Conclusions:

• The enhanced BED-YOLO model significantly improves tomato leaf disease detection accuracy, recall, and robustness.
• This method is highly suitable for intelligent disease monitoring in large-scale agricultural settings.
• The improvements make it a valuable tool for protecting tomato crop yield and quality.