Improving surface defect detection with context-guided asymmetric modulation networks and confidence-boosting loss

Danqing Kang, Jianhuang Lai, Yu Han

Abstract: Highlights•Lightweight but unexpectedly effective two-stage refinement framework.•Confidence-boosting loss can further improve performance gains.•Coarse predictions provide rich context for fine segmentation.•Efficient asymmetric modulation module drastically reduces calculations.