DyRAD: Radar Novel View Synthesis for Dynamic Driving Scenes
DyRAD renders dynamic radar range-azimuth-Doppler views, recovering 90.7% of RADIal detections versus 26.9%.
DyRAD reconstructs dynamic driving scenes for radar novel-view synthesis by combining static background reflectors with motion-tracked dynamic point reflectors and rendering full range-azimuth-Doppler tensors. Reflector velocities come from object tracks and are projected onto the line of sight, so Doppler supervises those tracks as well as appearing in the output. A fixed analytic point-spread function from the radar processing chain prevents sensor spread from being absorbed into scene geometry and enables zero-shot transfer to other radar configurations. On RADIal, DyRAD recovered detections for 90.7% of reference-detected objects, compared with 26.9% for the strongest baseline, and was also tested on Boreas and a synthetic benchmark.
- Models static background reflectors and tracked dynamic point reflectors.
- Doppler is both a rendered output and supervision for object tracks.
- A fixed analytic point-spread function keeps sensor spread out of geometry.
- The same scene can transfer across radar configurations without refitting.
- On RADIal it recovers 90.7% of detections versus 26.9% for the best baseline.
Full article225 words · extracted from huggingface.co · click to collapse
Reconstructing dynamic driving scenes from recorded sensor data supports closed-loop evaluation of autonomous driving systems by synthesizing observations beyond the original trajectory. Unlike cameras and LiDAR, radar measures radial velocity directly through Doppler. Yet existing radar novel-view synthesis fails to exploit this capability: methods addressing dynamic scenes reconstruct only range-azimuth tensors, while methods that render Doppler assume static scenes. Moreover, because radar processing spreads each reflection across multiple bins, existing representations absorb this spread into scene geometry, causing it to render incorrectly when the viewpoint moves. We present DyRAD, which models dynamic driving scenes using static background reflectors and motion-tracked dynamic point reflectors to render complete range-azimuth-Doppler (RAD) tensors. Reflector velocities are derived from object tracks and projected onto the line of sight, making Doppler both a rendered output and supervision for those tracks. Crucially, we render reflectors through a fixed analytic point-spread function (PSF) derived from the radar's signal-processing chain, preventing sensor-induced spread from being baked into the scene representation. Beyond improving scene reconstruction, this separation also enables zero-shot sensor-configuration transfer, allowing the same reconstructed scene to be rendered under different radar specifications without refitting. We evaluate DyRAD on RADIal, Boreas, and a synthetic benchmark across both on-path poses and displaced viewpoints untested by prior work. On RADIal, DyRAD recovers radar detections in 90.7% of reference-detected objects, compared with 26.9% for the strongest baseline.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.39841