DisTorch: A fast GPU implementation of 3D Hausdorff Distance

12 Apr 2025 (modified: 12 Apr 2025)MIDL 2025 Short Papers SubmissionEveryoneRevisionsBibTeXCC BY 4.0
Keywords: GPU accelerated, Hausdorff distance, 3D metrics
Abstract: In this short paper, we present a simple yet fast implementation of the Hausdorff distance, leveraging the KeOps package. After briefly describing the core ideas of the implementation, we compare the runtime and memory cost of different implementations, on different 3D volumes from different tasks. Most notably, we achieve sub 50ms per 3D scan across tasks, while having a much simpler codebase than the fastest alternative. Our implementation is freely available under the BSD license: https://github.com/jeromerony/distorch.
Submission Number: 108
Loading

OpenReview is a long-term project to advance science through improved peer review with legal nonprofit status. We gratefully acknowledge the support of the OpenReview Sponsors. © 2025 OpenReview