Unifi3D: A Study on 3D Representations for Generation and Reconstruction in a Common Framework
Nina Wiedemann, Sainan Liu, Quentin Leboutet, Katelyn Gao, Benjamin Ummenhofer Michael Paulitsch, Kai Yuan Intel Corporation · 2025
Read Unifi3D: A Study on 3D Representations for Generation and Reconstruction in a Common Framework by Nina Wiedemann, Sainan Liu, Quentin Leboutet, Katelyn Gao, Benjamin Ummenhofer Michael Paulitsch, Kai Yuan Intel Corporation (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-2509-02474.txt.
Description
Nina Wiedemann and six co-authors present a 2025 computer graphics study comparing ways to represent three-dimensional content for generation and reconstruction. Unlike images, which settled on pixels, 3D work still splits among voxel grids, point clouds, distance functions, and other formats.
Their Unifi3D framework scores each representation on quality, computing cost, and generalisation inside one shared pipeline. Steps from preprocessing through mesh rebuilding, compression, and generation are all tested, not just final models. A central finding is that reconstruction errors drag down overall results, so the two tasks should be judged together. The authors distil practical guidance for picking a representation by application. Code for the framework is released openly for reuse.
Catalog record
| CREATOR | Nina Wiedemann, Sainan Liu, Quentin Leboutet, Katelyn Gao, Benjamin Ummenhofer Michael Paulitsch, Kai Yuan Intel Corporation |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 69 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-02474