Neural energy regularization smooths LiDAR maps without extra inference cost
Original title: Neural Implicit Energy Optimization for LiDAR Mapping
Curvature energy on neural signed-distance level sets reduces jitter on million-point maps while keeping geometric fidelity.
Researchers formulate neural LiDAR mapping as optimization of surface-energy functionals over neural implicit signed distance fields. The regularizer uses the shape operator of the level set to penalize curvature and encourage smoother surfaces from sparse, noisy scans. Experiments on large-scale datasets report smoother reconstructions with preserved geometric fidelity and no added inference-time cost. Code is public at github.com/HrishikeshVish/nieo.
- The method targets artifacts in neural SDFs built from sparse, noisy LiDAR.
- Smoothness comes from an energy defined on the level-set shape operator.
- The authors say the approach scales to million-point LiDAR maps.
- They report no additional cost at inference time.
From the original article
1 min read at arxiv.orgNeural implicit representations provide memory-efficient signed distance field (SDF) parameterizations for large-scale LiDAR scenes, but the geometry recovered from sparse and often noisy inputs carries jitter and artifacts. Classical geometric optimization techniques resolve such artifacts with surface energy functionals, whose minimizers correspond to smooth surfaces.
By Hrishikesh Viswanath · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.