TL;DR: A multi-scale, point-based surface reconstruction method that leverages point cloud serialization for fast neighbor approximation.

Note: Reconstruction results of Carla subset.

Paper

@misc{li2025noksrkernelfreeneuralsurface,
                    title={NoKSR: Kernel-Free Neural Surface Reconstruction via Point Cloud Serialization}, 
                    author={Zhen Li and Weiwei Sun and Shrisudhan Govindarajan and Shaobo Xia and Daniel Rebain and Kwang Moo Yi and Andrea Tagliasacchi},
                    year={2025},
                    eprint={2502.12534},
                    archivePrefix={arXiv},
                    primaryClass={cs.CV},
                    url={https://arxiv.org/abs/2502.12534}, 
              }
                

Abstract

We present a novel approach to large-scale point cloud surface reconstruction by developing an efficient framework that converts an irregular point cloud into a signed distance field (SDF). Our backbone builds upon recent transformer- based architectures (i.e. PointTransformerV3), that serial- izes the point cloud into a locality-preserving sequence of tokens. We efficiently predict t he SDF value at a point by ag- gregating nearby tokens, where fast approximate neighbors can be retrieved thanks to the serialization. We serialize the point cloud at different levels/scales, and non-linearly aggregate a feature to predict the SDF value. We show that aggregating across multiple scales is critical to over- come the approximations introduced by the serialization (i.e. false negatives in the neighborhood). Our frameworks sets the new state-of-the-art in terms of accuracy and effi- ciency (better or similar performance with half the latency of the best prior method, coupled with a simpler implemen- tation), particularly on outdoor datasets where sparse-grid methods have shown limited performance.

Method

Overview: We map a sparse input point cloud with a point cloud backbone into a point feature hierarchy, from which we compute the signed distance of a query. At each level, we utilize the efficient procedure to retrieve local neighborhoods of the query. We then compute per-level features with the aggregation module. At last, we sum per-level features and convert it into the signed distance with an MLP.

Neighborhood function: (left) retrieving a local neighborhood with K-nearest neighbor(KNN) or ball-query methods is challenging to implement efficiently on GPU hardware. (right) we propose to retrieve a neighborhood from a 1-D ordered list, by serializing points along a Hibert curve, and excluding the impact of points distant from the query (i.e. remove false positives).

NKSR vs NoKSR(ours)

We compare NoKSR(ours) with baselines including Neural Kernel Surface Reconstruction (NKSR). Our method achieves high quality surface reconstructions which preserve more details than NKSR, which loses information due to quantization for large and non-uniformly sampled datasets like Carla. Note that we can move slider left and right to see the comparison.

More Reconstruction Results

Outdoor Scene: Carla

Indoor Scene: ScanNet

Indoor Scene: SyntheticRoom

Acknowledgements