A Glove-Wearing Detection Algorithm Based on Improved YOLOv8

Area of Science:

• Computer Vision
• Machine Learning
• Occupational Safety

Background:

• Accidental injuries in workshops, including mechanical damage and burns, are a significant concern during machinery operation.
• Ensuring workers wear appropriate personal protective equipment, such as gloves, is crucial for accident prevention.

Purpose of the Study:

• To develop and evaluate an efficient and accurate glove detection algorithm for workshop environments.
• To improve upon existing object detection models for enhanced safety in industrial settings.

Main Methods:

• Proposed a novel glove detection algorithm, YOLOv8-AFPN-M-C2f, based on the YOLOv8 architecture.
• Innovated by replacing the YOLOv8 head with the AFPN-M-C2f network to improve feature vector propagation and reduce semantic discrepancies.
• Incorporated a superficial feature layer to enhance detection of smaller objects and surface features.

Main Results:

• The YOLOv8-AFPN-M-C2f model demonstrated superior performance compared to other network models on a factory glove detection dataset.
• Achieved a 2.6% increase in mean Average Precision at 50% IoU (mAP@50%), a 63.8% increase in Frames Per Second (FPS), and a 13% reduction in parameters compared to the baseline YOLOv8.
• The enhanced model showed improved sensitivity to smaller objects and richer surface feature information.

Conclusions:

• The YOLOv8-AFPN-M-C2f algorithm provides an effective solution for real-time glove detection in workshop settings.
• The proposed modifications significantly enhance detection speed, accuracy, and computational efficiency, contributing to improved occupational safety.