YOLO-PLNet: a lightweight real-time detection model for peanut leaf diseases based on edge deployment

Area of Science:

• Agricultural technology
• Computer vision
• Deep learning

Background:

• Peanut leaf diseases significantly impact crop yield and quality.
• Early and accurate disease detection is crucial for effective management.
• Existing lightweight deep learning models face challenges in balancing size, accuracy, and edge deployment for agricultural applications.

Purpose of the Study:

• To propose YOLO-PLNet, a novel lightweight real-time detection model for edge deployment.
• To improve the accuracy and efficiency of peanut leaf disease detection in complex agricultural environments.
• To enhance the adaptability of deep learning models for large-scale agricultural monitoring on edge devices.

Main Methods:

• Developed YOLO-PLNet based on YOLO11n with a lightweight backbone and Neck structure.
• Incorporated a Lightweight Attention-Enhanced (LAE) convolution module and a Channel-Spatial Attention Mechanism (CBAM) for improved feature representation.
• Utilized an Asymptotic Feature Pyramid Network (AFPN) with staged cross-level fusion for multi-scale detection.

Main Results:

• YOLO-PLNet achieved significant reductions in parameter count (18.07%), computational complexity (16.92%), and model size (15.70%) compared to the baseline.
• The model reached high detection accuracy with mAP@0.5 of 98.1% and mAP@0.5:0.95 of 94.7%.
• On the Jetson Orin NX platform, YOLO-PLNet demonstrated real-time detection speeds (up to 41.3 FPS at INT8 precision) with reduced latency, GPU usage, and power consumption.

Conclusions:

• YOLO-PLNet offers a highly accurate and efficient solution for real-time peanut leaf disease detection on edge devices.
• The model's lightweight design and enhanced feature representation improve adaptability for complex agricultural scenarios.
• This provides a feasible approach for intelligent, large-scale monitoring of peanut crop health.