LG-FA incrementally builds a sparse global vector layer from per-frame BEV predictions, uses class-constrained alignment to refine ego pose and complete topology, then reprojects the augmented foreground to improve per-frame outputs.
Streammapnet: Streaming mapping network for vectorized online hd map construction
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Localization-Guided Foreground Augmentation in Autonomous Driving
LG-FA incrementally builds a sparse global vector layer from per-frame BEV predictions, uses class-constrained alignment to refine ego pose and complete topology, then reprojects the augmented foreground to improve per-frame outputs.