MoRe: Motion-aware Feed-forward 4D Reconstruction Transformer

Junton Fang, Zequn Chen, Weiqi Zhang, Donglin Di, Xuancheng Zhang, Chengmin Yang, Yu-Shen Liu · 2026

Read MoRe: Motion-aware Feed-forward 4D Reconstruction Transformer by Junton Fang, Zequn Chen, Weiqi Zhang, Donglin Di, Xuancheng Zhang, Chengmin Yang, Yu-Shen Liu (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-2603-05078.txt.

Description

Juntong Fang and team offer MoRe, a feedforward 4D rebuild network recovering dynamic 3D scenes from monocular videos. Moving objects corrupt camera pose estimation, and current tuning methods need extra supervision while remaining computationally expensive and impractical in real time. Built on a strong static rebuild backbone, MoRe uses attention-forcing to disentangle dynamic motion from static structure, plus grouped causal attention capturing time dependencies across varying token lengths. Fine-tuning on large-scale diverse data sets adds toughness, and tests show high-quality dynamic reconstructions with exceptional leanness.

Catalog record

CREATORJunton Fang, Zequn Chen, Weiqi Zhang, Donglin Di, Xuancheng Zhang, Chengmin Yang, Yu-Shen Liu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43683
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.05078

Browse the full library · About the Tower of Babel