AI
DyRAD: Radar Novel View Synthesis for Dynamic Driving Scenes
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 29/09/2026, 20:00:00 UTC
Briefing created: 01/10/2026, 09:35:02 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 01/10/2026, 09:18:24 UTC.
Briefings
DyRAD is a novel view synthesis method for dynamic driving scenes that models reflectors as static background objects and motion-tracked dynamic point reflectors. By separating radar signal processing from scene representation, DyRAD renders complete range-azimuth-Doppler tensors without sensor-induced spread errors. Evaluated on RADIal, DyRAD recovers 90.7% of reference-detected objects, significantly outperforming the strongest baseline at 26.9%. The method enables zero-shot sensor-configuration transfer, allowing the same reconstruction to work with different radar specifications.
Key points
- DyRAD models dynamic driving scenes using static background reflectors and motion-tracked dynamic point reflectors to render complete range-azimuth-Doppler tensors.
- DyRAD recovers radar detections in 90.7% of reference-detected objects on the RADIal benchmark, compared with 26.9% for the strongest baseline.
Context and limitations
Evaluation is limited to RADIal, Boreas, and a synthetic benchmark across on-path poses and displaced viewpoints untested by prior work. Doppler velocities are derived from object tracks rather than direct sensor measurements, and the fixed analytic PSF prevents modeling of sensor-induced spread variations.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 01/10/2026, 09:35:02 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeibjcvaikejfmsyfcredwikpmnaqv45beirwawfvpi2aisz5zrkvkm