LPFF-DETR: a lightweight transformer model for printed circuit board defect detection

Accurate detection of minute defects in printed circuit boards (PCBs) is crucial for quality control, yet remains challenging due to the small size of defects and the complexity of background textures. Conventional detection methods often struggle to balance detection accuracy and computational efficiency, which limits their applicability in resource-constrained industrial scenarios. To address this issue, we propose lightweight PCB feature fusion detection transformer (LPFF-DETR), a lightweight DETR-based framework for PCB defect detection that prioritizes computational efficiency as a primary design objective. Specifically, a lightweight feature extraction architecture is designed to reduce model complexity while improving the representation of small and subtle defects, and an efficient multiscale feature fusion strategy is developed to enhance cross-scale information interaction without introducing substantial computational overhead. Experimental results on the HRIPCB dataset show that LPFF-DETR achieves an mAP (Formula presented) (Formula presented) of 98.5% with only 7.7 M parameters and 11.2 GFLOPs, outperforming existing detectors that require significantly higher computational resources. These results demonstrate that the proposed method achieves a favorable balance between detection performance and computational efficiency, making it particularly suitable for resource-constrained industrial PCB inspection scenarios.