EvalMVX: A Unified Benchmarking for Neural 3D Reconstruction under Diverse Multiview Setups

Zaiyan Yang 1 Jieji Ren 2 Xiangyi Wang 1 Zonglin Li 1 Xu Cao 3 Heng Guo 1,4∗ Zhanyu Ma 1 Boxin Shi 5 1 Beijing University of Posts and Telecommunications, Beijing, China 2 School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China 3 Independent Researcher 4 Xiong’an Aerospace I · 2026

Read EvalMVX: A Unified Benchmarking for Neural 3D Reconstruction under Diverse Multiview Setups by Zaiyan Yang 1 Jieji Ren 2 Xiangyi Wang 1 Zonglin Li 1 Xu Cao 3 Heng Guo 1,4∗ Zhanyu Ma 1 Boxin Shi 5 1 Beijing University of Posts and Telecommunications, Beijing, China 2 School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China 3 Independent Researcher 4 Xiong’an Aerospace I (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-2602-24065.txt.

Description

Zaiyan Yang and seven co-authors introduce EvalMVX, a unified benchmark for neural three-dimensional reconstruction under diverse multiview setups. The 2026 paper notes that current real-world datasets mainly benchmark multiview stereo from color images. Multiview photometric stereo and multiview shape from polarization, both vital for high-fidelity surfaces and sparse inputs, have never been assessed together with it.

EvalMVX contains many objects, each photographed with a polarized camera under many views and lighting conditions, including one-light-at-a-time and natural illumination, yielding thousands of images. Each object carries an aligned ground-truth mesh for quantitative benchmarking across all three techniques at once. The authors evaluate many recent methods, record the best performers, and identify open problems across varied geometry and reflectance. The benchmark aims to guide research toward reconstruction methods that work across capture setups, not just one.

Catalog record

CREATORZaiyan Yang 1 Jieji Ren 2 Xiangyi Wang 1 Zonglin Li 1 Xu Cao 3 Heng Guo 1,4∗ Zhanyu Ma 1 Boxin Shi 5 1 Beijing University of Posts and Telecommunications, Beijing, China 2 School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai, China 3 Independent Researcher 4 Xiong’an Aerospace I
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-24065

Browse the full library · About the Tower of Babel