BEV-Denoise applies a DDPM-inspired UNet to estimate and remove intrinsic noise from BEV features, improving semantic segmentation when added to existing view transformation models on nuScenes.
Title resolution pending
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
-
BEV-Denoise: Learning Intrinsic Noise for Accurate Bird's-Eye-View Semantic Segmentation
BEV-Denoise applies a DDPM-inspired UNet to estimate and remove intrinsic noise from BEV features, improving semantic segmentation when added to existing view transformation models on nuScenes.