YOLOv8-DEE: a high-precision model for printed circuit board defect detection.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Manufacturing Technology

Background:

• Printed circuit board (PCB) defects impact consumer electronics quality, stability, and reliability.
• Existing PCB defect inspection models face accuracy limitations due to complex backgrounds and multi-scale defect characteristics.

Purpose of the Study:

• To introduce an enhanced YOLOv8 network, YOLOv8-DEE, for improved accuracy in detecting printed circuit board (PCB) defects.
• To address the challenges of complex backgrounds and multi-scale features in PCB defect detection.

Main Methods:

• Developed YOLOv8-DEE by enhancing the YOLOv8-L model with depthwise separable convolution (DSC) modules in the backbone for superior feature extraction.
• Integrated an efficient multi-scale attention (EMA) module in the network's neck to improve contextual information interaction.
• Replaced the complete intersection over union (CIoU) loss with efficient intersection over union (EIoU) loss for better localization of defects with varying sizes and aspect ratios.

Main Results:

• YOLOv8-DEE achieved a mean average precision (mAP) of 97.5% on the HRIPCB dataset and 98.7% on the DeepPCB dataset.
• Demonstrated performance improvements of 2.5% and 0.7% over the baseline YOLOv8-L model on respective datasets.
• Outperformed other state-of-the-art methods in detecting small, medium, and large PCB defects.

Conclusions:

• The proposed YOLOv8-DEE model effectively enhances PCB defect detection accuracy and reliability.
• The integration of DSC, EMA, and EIoU modules contributes to superior performance in complex industrial inspection scenarios.
• YOLOv8-DEE represents a significant advancement in automated visual inspection for electronic manufacturing.