Sparse Code Uplifting for Efficient 3D Language Gaussian Splatting

Lovre Antonio Budimir, Yushi Guan, Steve Ryhner, Sven Lončarić, Nandita Vijaykumar, [0.8em] Faculty of Electrical Engineering, Computing, University of Zagreb · 2026

Read Sparse Code Uplifting for Efficient 3D Language Gaussian Splatting by Lovre Antonio Budimir, Yushi Guan, Steve Ryhner, Sven Lončarić, Nandita Vijaykumar, [0.8em] Faculty of Electrical Engineering, Computing, University of Zagreb (2026) 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-2605-13600.txt.

Description

Lovre Antonio Budimir and four colleagues propose SCOUP, sparse code uplifting for efficient 3D language Gaussian splatting. 3D language Gaussian splatting adds word features to 3D scenes for open-vocabulary understanding. SCOUP learns sparse codebook representations from 2D image regions alone, then uplifts them to 3D Gaussians by weighted sparse aggregation. Top-K filtering keeps the dominant multi-view coefficients per Gaussian, giving cheap storage and fast rendering. It trains up to 400 times faster with 3 times less memory, matching or beating rivals on open-vocabulary querying.

Catalog record

CREATORLovre Antonio Budimir, Yushi Guan, Steve Ryhner, Sven Lončarić, Nandita Vijaykumar, [0.8em] Faculty of Electrical Engineering, Computing, University of Zagreb
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43865
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.13600

Browse the full library · About the Tower of Babel