SymmCompletion: High-Fidelity and High-Consistency Point Cloud Completion with Symmetry Guidance

Hongyu Yan, Zijun Li, Kunming Luo, Li Lu, Ping Tan · 2025

Read SymmCompletion: High-Fidelity and High-Consistency Point Cloud Completion with Symmetry Guidance by Hongyu Yan, Zijun Li, Kunming Luo, Li Lu, Ping Tan (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-2503-18007.txt.

Description

Hongyu Yan, Zijun Li, Kunming Luo, and a wider team wrote this computer vision paper, published in 2025. It completes partial 3D scans by exploiting symmetry, mirroring what a scanner saw on one side to rebuild what it missed on the other.

Most completion methods get the overall shape right but blur fine detail, and the rebuilt parts clash with the scanned parts. The new system works in two stages. A first network estimates, point by point, the local mirror transform, flipping key geometry into the gaps to draft a full shape. A transformer then refines that draft, guided explicitly by the matched partial-missing pairs. The guidance ties old and new geometry together. On standard benchmarks, the method beat leading rivals in both looks and numbers, producing detailed, consistent shapes.

Catalog record

CREATORHongyu Yan, Zijun Li, Kunming Luo, Li Lu, Ping Tan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-18007

Browse the full library · About the Tower of Babel