MGL-YOLO: A Lightweight Barcode Target Detection Algorithm.

Area of Science:

• Computer Vision
• Machine Learning
• Deep Learning

Background:

• One-dimensional barcode detection is crucial for logistics, retail, and manufacturing efficiency.
• Deploying deep convolutional neural networks on edge devices is challenging due to resource limitations.

Purpose of the Study:

• To propose MGL-YOLO, a lightweight one-dimensional barcode detection network.
• To achieve high detection accuracy with low computational cost for embedded systems.

Main Methods:

• Introduced a multi-scale group convolution (MSGConv) integrated into the C2f module (MSG-C2f) for enhanced multi-scale feature extraction.
• Designed the Group RepConv Cross Stage Partial Efficient Long-Range Attention Network (GRCE) to optimize feature extraction in the neck section.
• Proposed a Lightweight Shared Multi-Scale Detection Head (LSMD) to reduce parameters and complexity.

Main Results:

• MGL-YOLO improved MAP50 by 2.57% and MAP50.95 by 2.31% compared to YOLOv8.
• Reduced parameter size by 36.21% and computational cost by 34.15%.
• Demonstrated superior average precision against other object detection networks.

Conclusions:

• MGL-YOLO effectively addresses the challenges of one-dimensional barcode detection on resource-constrained devices.
• The proposed network achieves high accuracy and efficiency, making it suitable for real-world applications.
• MGL-YOLO offers a promising solution for improving operational efficiency in various industries.