LexiCan: Lexicographic Canonicalization of Point Clouds for Machine Learning Applications

Name: LexiCan: Lexicographic Canonicalization of Point Clouds for Machine Learning Applications
Uploaded: 2026-10-18T22:19:37.125249+00:00
Duration: 390 s
Description: LexiCan is a new method for labeling 3D point clouds that creates a provable canonical representative invariant to rotations (SO(3) or O(3)) and permutations. It uses sorting based on label rarity ranks to ensure high-precision invariance, avoiding the jumps seen in older methods like PCA and ASUN, leading to better performance for non-invariant models.

LexiCan is a new method for labeling 3D point clouds that creates a provable canonical representative invariant to rotations (SO(3) or O(3)) and permutations. It uses sorting based on label rarity ranks to ensure high-precision invariance, avoiding the jumps seen in older methods like PCA and ASUN, leading to better performance for non-invariant models.

Open the full episode