Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SME-YOLO: A Real-Time Detector for Tiny Defect Detection on PCB Surfaces

Meng Han

TL;DR

This work tackles the challenge of real-time detection of extremely tiny PCB surface defects by extending YOLOv11n into SME-YOLO, a lightweight detector that integrates three targeted innovations: NWDLoss, which models bounding boxes as 2D Gaussians and uses Normalized Wasserstein Distance to stabilize localization of small objects; MSFA, a scale-focused attention module that concentrates on the dominant defect scales with an implicit ensemble design; and EUCB, an efficient upsampling block that preserves edge and texture details during feature map reconstruction. The approach yields state-of-the-art results on the PKU-Market-PCB dataset, notably improving mAP and precision over baselines while maintaining low computational overhead. The combination of robust tiny-object localization, focused multi-scale feature extraction, and efficient upsampling enables reliable real-time PCB defect detection, with implications for manufacturing quality control. Future work includes evaluating on larger, more diverse datasets and exploring robustness under challenging imaging conditions.

Abstract

Surface defects on Printed Circuit Boards (PCBs) directly compromise product reliability and safety. However, achieving high-precision detection is challenging because PCB defects are typically characterized by tiny sizes, high texture similarity, and uneven scale distributions. To address these challenges, this paper proposes a novel framework based on YOLOv11n, named SME-YOLO (Small-target Multi-scale Enhanced YOLO). First, we employ the Normalized Wasserstein Distance Loss (NWDLoss). This metric effectively mitigates the sensitivity of Intersection over Union (IoU) to positional deviations in tiny objects. Second, the original upsampling module is replaced by the Efficient Upsampling Convolution Block (EUCB). By utilizing multi-scale convolutions, the EUCB gradually recovers spatial resolution and enhances the preservation of edge and texture details for tiny defects. Finally, this paper proposes the Multi-Scale Focused Attention (MSFA) module. Tailored to the specific spatial distribution of PCB defects, this module adaptively strengthens perception within key scale intervals, achieving efficient fusion of local fine-grained features and global context information. Experimental results on the PKU-PCB dataset demonstrate that SME-YOLO achieves state-of-the-art performance. Specifically, compared to the baseline YOLOv11n, SME-YOLO improves mAP by 2.2% and Precision by 4%, validating the effectiveness of the proposed method.

SME-YOLO: A Real-Time Detector for Tiny Defect Detection on PCB Surfaces

TL;DR

This work tackles the challenge of real-time detection of extremely tiny PCB surface defects by extending YOLOv11n into SME-YOLO, a lightweight detector that integrates three targeted innovations: NWDLoss, which models bounding boxes as 2D Gaussians and uses Normalized Wasserstein Distance to stabilize localization of small objects; MSFA, a scale-focused attention module that concentrates on the dominant defect scales with an implicit ensemble design; and EUCB, an efficient upsampling block that preserves edge and texture details during feature map reconstruction. The approach yields state-of-the-art results on the PKU-Market-PCB dataset, notably improving mAP and precision over baselines while maintaining low computational overhead. The combination of robust tiny-object localization, focused multi-scale feature extraction, and efficient upsampling enables reliable real-time PCB defect detection, with implications for manufacturing quality control. Future work includes evaluating on larger, more diverse datasets and exploring robustness under challenging imaging conditions.

Abstract

Surface defects on Printed Circuit Boards (PCBs) directly compromise product reliability and safety. However, achieving high-precision detection is challenging because PCB defects are typically characterized by tiny sizes, high texture similarity, and uneven scale distributions. To address these challenges, this paper proposes a novel framework based on YOLOv11n, named SME-YOLO (Small-target Multi-scale Enhanced YOLO). First, we employ the Normalized Wasserstein Distance Loss (NWDLoss). This metric effectively mitigates the sensitivity of Intersection over Union (IoU) to positional deviations in tiny objects. Second, the original upsampling module is replaced by the Efficient Upsampling Convolution Block (EUCB). By utilizing multi-scale convolutions, the EUCB gradually recovers spatial resolution and enhances the preservation of edge and texture details for tiny defects. Finally, this paper proposes the Multi-Scale Focused Attention (MSFA) module. Tailored to the specific spatial distribution of PCB defects, this module adaptively strengthens perception within key scale intervals, achieving efficient fusion of local fine-grained features and global context information. Experimental results on the PKU-PCB dataset demonstrate that SME-YOLO achieves state-of-the-art performance. Specifically, compared to the baseline YOLOv11n, SME-YOLO improves mAP by 2.2% and Precision by 4%, validating the effectiveness of the proposed method.
Paper Structure (27 sections, 11 equations, 8 figures, 5 tables)

This paper contains 27 sections, 11 equations, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. 1)Introduction of Normalized Wasserstein Distance Loss (NWDLoss).
  3. 2)Optimization of upsampling with EUCB.
  4. 3)Enhancement of multi-scale feature extraction with MSFA.
  5. Relate work
  6. Overview of PCB Surface Defect Detection Methods
  7. YOLO-based PCB Defect Detection
  8. Methodology
  9. Overall Architecture of SME-YOLO
  10. Improved Content Description
  11. NWDLoss: Normalized Wasserstein Distance Loss
  12. MSFA: Multi-Scale Focused Attention
  13. Scale Focusing:
  14. Implicit Ensemble:
  15. Computational Efficiency Optimization:
  16. ...and 12 more sections

Figures (8)

  • Figure 1: SME-YOLO Overall Architecture.
  • Figure 2: Comparison of IoU sensitivity to misalignment for tiny targets and for conventional-sized targets.
  • Figure 3: Overall Structure of the MSFA Module.
  • Figure 4: Flowchart of the EUCB Module.
  • Figure 5: Six Types of PCB Surface Defects.
  • ...and 3 more figures