DiffTrans: Differentiable Geometry-Materials Decomposition for Reconstructing Transparent Objects

Changpu Li, Shuang Wu, Songlin Tang, Guangming Lu, Jun Yu, Wenjie Pei · 2026

Read DiffTrans: Differentiable Geometry-Materials Decomposition for Reconstructing Transparent Objects by Changpu Li, Shuang Wu, Songlin Tang, Guangming Lu, Jun Yu, Wenjie Pei (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-2603-00413.txt.

Description

Li and five colleagues rebuild transparent objects from multi view photos, a hard task because light bends and scatters in complex ways. Their DiffTrans framework first shapes a rough geometry from object silhouettes with a flexible surface model, then recovers scene lighting. A recursive ray tracer then refines shape, refraction, and absorption together in one end to end pass, running fast on graphics hardware. Tests on several benchmarks beat rival methods, most clearly in tangled scenes with varied shapes and textures.

Catalog record

CREATORChangpu Li, Shuang Wu, Songlin Tang, Guangming Lu, Jun Yu, Wenjie Pei
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63009
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2603.00413

Browse the full library · About the Tower of Babel