PseudoMapLabeler Uses Confidence-Aware Refinement for HD Map Construction
August 14, 2026
PseudoMapLabeler utilizes a teacher-student semi-supervised framework to generate high-quality pseudo-labels for online mapping. It employs Beta-distribution-based confidence maps and spatial clipping to selectively preserve reliable map elements from unlabeled data.
HOW THIS AFFECTS YOU
●
researcherThis introduces a more granular method for refining pseudo-labels in semi-supervised spatial learning.