Python
Aim:
To develop a lightweight, accurate, and efficient YOLO-based deep learning model for detecting and classifying pavement defects such as cracks and potholes in real time, optimized for deployment.
Abstract:
This project introduces LMD_YOLO, a lightweight and efficient deep learning model for pavement defect detection. Traditional methods struggle to balance accuracy, speed, and computational cost, especially under diverse road and lighting conditions. LMD_YOLO addresses these challenges through the integration of key innovations: the detection head, the lightweight convolutional module, and an optimized LeYOLO backbone. These enhancements enable the model to achieve higher accuracy and faster inference while maintaining reduced resource consumption. Experimental results on the demonstrate superior mean Average Precision, validating its capability for real-time road surface analysis.
Proposed System:
The proposed LMD_YOLO model integrates YOLOv11n with architectural enhancements to achieve high efficiency and precision. Key improvements include replacing the standard backbone with the LeYOLO structure for reduced computation, incorporating the module in the neck network to improve feature fusion in the detection head for enhanced feature representation.This optimized model achieves superior accuracy in detecting multiple types of pavement defects, including longitudinal cracks, transverse cracks, alligator cracks, potholes and background while ensuring faster real-time processing suitable for edge deployment.
Advantage:
- Enhanced detection accuracy for multi-scale pavement defect.
- Reduced computational cost and improved real-time performance.
Robust detection under varying lighting and environmental conditions.
Reviews
There are no reviews yet.