Garlic-YOLO-DD: a lightweight object detection algorithm for garlic damage detection.

Area of Science:

• Computer Vision
• Agricultural Technology
• Machine Learning

Background:

• Resource-constrained environments pose challenges for existing garlic damage detection models.
• High computational complexity and excessive parameters limit real-time applications of current methods.

Purpose of the Study:

• To develop a lightweight and efficient object detection algorithm for automated garlic damage recognition.
• To address the limitations of existing models in terms of computational load and parameter count.

Main Methods:

• Proposed Garlic-YOLO-DD, a lightweight single-stage object detection algorithm based on YOLOv11n.
• Implemented ADown module to reduce parameters and computational load in the backbone network.
• Integrated SimAM attention mechanism for enhanced feature extraction of subtle lesions.
• Utilized BiFPN architecture for optimized multi-scale feature fusion.

Main Results:

• Garlic-YOLO-DD reduced parameters to 57.96% of YOLOv11n.
• Computational load decreased by 20.63%, while inference speed increased by 15.97%.
• Achieved a mean Average Precision (mAP@50%) of 27.64% on a self-built dataset.

Conclusions:

• Garlic-YOLO-DD offers an efficient and accurate solution for real-time garlic damage detection.
• The model is suitable for deployment in intelligent agricultural systems with limited computational resources.