FA-KPConv: Introducing Euclidean Symmetries to KPConv via Frame Averaging
Ali Alawieh 1 2 and Alexandru P. Condurache 1 2 · 2025
Read FA-KPConv: Introducing Euclidean Symmetries to KPConv via Frame Averaging by Ali Alawieh 1 2 and Alexandru P. Condurache 1 2 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2505-04485.txt.
Description
Ali Alawieh and Alexandru Condurache give point cloud networks exact respect for Euclidean symmetry, using frame averaging over the KPConv backbone. KPConv is a workhorse for 3D point cloud analysis, yet it only learns invariance to shifts, turns, and flips roughly, from heavy training data. Exact symmetry has to be built in, not hoped for.
Their 2025 paper wraps an existing KPConv network in a frame averaging shell that can be set for invariance, equivariance, or a mix, per motion type. The wrap adds no learnable parameters and drops no input detail. Gains show most where data is scarce or test clouds arrive randomly rotated, in both classification and registration. The bias does the work that augmentation used to fake.
Catalog record
| CREATOR | Ali Alawieh 1 2 and Alexandru P. Condurache 1 2 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-04485