YOLO-SDA: an innovative YOLOv12-derived model with superior performance in recognizing peanut foliar diseases

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Manual detection of peanut leaf diseases causes significant time lags, leading to widespread epidemics and yield losses.
• Precise and rapid detection is crucial for mitigating large-scale disease outbreaks in peanut production.
• Optimized detection algorithms are essential for intelligent monitoring equipment.

Purpose of the Study:

• To develop an improved deep learning model for efficient and accurate peanut leaf disease detection.
• To enhance the YOLOv12 algorithm's performance and reduce its computational burden for practical deployment.

Main Methods:

• The YOLOv12 algorithm was used as the baseline model.
• Three modules—StarNet, DySample, and A2C2f_SCSA—were integrated into YOLOv12 to create the YOLO-SDA model.
• The integrated modules aim to optimize feature extraction, sampling efficiency, and feature fusion.

Main Results:

• The YOLO-SDA model demonstrated superior performance and efficiency compared to YOLOv12.
• Achieved a 44% reduction in parameters, 38.5% decrease in GFLOPs, and 43.6% reduction in model size.
• Improved detection precision by 2.0% and mAP@0.5-0.95 by 2.5%.

Conclusions:

• The YOLO-SDA model effectively enhances peanut leaf disease detection accuracy and efficiency.
• Reduced computational requirements make YOLO-SDA suitable for resource-constrained, on-site monitoring equipment.
• The model provides reliable technical support for preventing disease outbreaks and protecting peanut yields.