Car-GS: Addressing Reflective and Transparent Surface Challenges in 3D Car Reconstruction

Congcong Li, Jin Wang, Xiaomeng Wang, Xingchen Zhou, Wei Wu, Yuzhi Zhang, Tongyi CaoDeepRoute.AI {congcongli, jinwang03, xiaomengwang, xingchenzhou, weiwu, yuzhizhang, tongyicao}@deeproute.ai · 2025

Read Car-GS: Addressing Reflective and Transparent Surface Challenges in 3D Car Reconstruction by Congcong Li, Jin Wang, Xiaomeng Wang, Xingchen Zhou, Wei Wu, Yuzhi Zhang, Tongyi CaoDeepRoute.AI {congcongli, jinwang03, xiaomengwang, xingchenzhou, weiwu, yuzhizhang, tongyicao}@deeproute.ai (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-2501-11020.txt.

Description

Congcong Li and co-authors tackle three-dimensional car reconstruction in a 2025 paper. Cars are hard to rebuild digitally because paint reflects light sharply and glass is both see-through and shiny. Standard scene methods blur these surfaces, so the team designs a method tuned to cars, named Car-GS in the paper.

Three fixes do the work. View-dependent primitives model reflections, each primitive gains its own geometry-only opacity for transparent parts, and a quality-aware module leans on a pretrained normal estimator where camera angles face glass head-on. Tests show cleaner surfaces and clear wins over prior methods. The authors target uses in self-driving, virtual reality, and games.

Catalog record

CREATORCongcong Li, Jin Wang, Xiaomeng Wang, Xingchen Zhou, Wei Wu, Yuzhi Zhang, Tongyi CaoDeepRoute.AI {congcongli, jinwang03, xiaomengwang, xingchenzhou, weiwu, yuzhizhang, tongyicao}@deeproute.ai
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-11020

Browse the full library · About the Tower of Babel