Rethinking Radar 3D Perception: Representation Learning from Raw Spectra
Rethinking Radar 3D Perception: Representation Learning from Raw Spectra
This paper challenges the assumption that raw radar spectra are too noisy for direct learning. The authors introduce Leakage-aware Spectral MAE, a masked autoencoder that uses corruption strategies inspired by spectral leakage to separate structured signal from noise without handcrafted reductions. By treating Doppler as a primary spatial dimension in an x-y-D representation, the method outperforms state-of-the-art radar-based detectors and point-cloud-based self-supervised learning counterparts on 3D perception tasks.
The authors propose a self-supervised masked autoencoder approach for raw radar spectra, arguing that structured signal can be separated from noise via reconstruction robustness rather than handcrafted reductions.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
Source：arXiv Robotics — research abstracts · arxiv.org