English

ASH: A Modern Framework for Parallel Spatial Hashing in 3D Perception

Computer Vision and Pattern Recognition 2023-01-31 v2 Distributed, Parallel, and Cluster Computing Graphics Robotics

Abstract

We present ASH, a modern and high-performance framework for parallel spatial hashing on GPU. Compared to existing GPU hash map implementations, ASH achieves higher performance, supports richer functionality, and requires fewer lines of code (LoC) when used for implementing spatially varying operations from volumetric geometry reconstruction to differentiable appearance reconstruction. Unlike existing GPU hash maps, the ASH framework provides a versatile tensor interface, hiding low-level details from the users. In addition, by decoupling the internal hashing data structures and key-value data in buffers, we offer direct access to spatially varying data via indices, enabling seamless integration to modern libraries such as PyTorch. To achieve this, we 1) detach stored key-value data from the low-level hash map implementation; 2) bridge the pointer-first low level data structures to index-first high-level tensor interfaces via an index heap; 3) adapt both generic and non-generic integer-only hash map implementations as backends to operate on multi-dimensional keys. We first profile our hash map against state-of-the-art hash maps on synthetic data to show the performance gain from this architecture. We then show that ASH can consistently achieve higher performance on various large-scale 3D perception tasks with fewer LoC by showcasing several applications, including 1) point cloud voxelization, 2) retargetable volumetric scene reconstruction, 3) non-rigid point cloud registration and volumetric deformation, and 4) spatially varying geometry and appearance refinement. ASH and its example applications are open sourced in Open3D (http://www.open3d.org).

Keywords

Cite

@article{arxiv.2110.00511,
  title  = {ASH: A Modern Framework for Parallel Spatial Hashing in 3D Perception},
  author = {Wei Dong and Yixing Lao and Michael Kaess and Vladlen Koltun},
  journal= {arXiv preprint arXiv:2110.00511},
  year   = {2023}
}

Comments

18 pages, 19 figures

R2 v1 2026-06-24T06:33:37.270Z