Robust 4D Visual Geometry Transformer with Uncertainty-Aware Priors
2026-04-10 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors work on reconstructing changing 3D scenes over time, which is hard because movement can confuse the process. They propose a new method that separates moving parts from still parts by carefully handling uncertainty during reconstruction. Their method uses three main ideas: focusing on important motion signals, keeping geometry consistent by checking neighbors, and using confidence levels to improve multi-view consistency. Tests show their approach is more accurate and better at segmenting objects than previous methods, without needing extra tuning per scene.
4D reconstructiondynamic scenesuncertainty modelingmulti-head attentionentropy-guided projectiongeometry purificationcross-view consistencyheteroscedastic maximum likelihooddepth confidencefeed-forward inference
Authors
Ying Zang, Yidong Han, Chaotao Ding, Yuanqi Hu, Deyi Ji, Qi Zhu, Xuanfu Li, Jin Ma, Lingyun Sun, Tianrun Chen, Lanyun Zhu
Abstract
Reconstructing dynamic 4D scenes is an important yet challenging task. While 3D foundation models like VGGT excel in static settings, they often struggle with dynamic sequences where motion causes significant geometric ambiguity. To address this, we present a framework designed to disentangle dynamic and static components by modeling uncertainty across different stages of the reconstruction process. Our approach introduces three synergistic mechanisms: (1) Entropy-Guided Subspace Projection, which leverages information-theoretic weighting to adaptively aggregate multi-head attention distributions, effectively isolating dynamic motion cues from semantic noise; (2) Local-Consistency Driven Geometry Purification, which enforces spatial continuity via radius-based neighborhood constraints to eliminate structural outliers; and (3) Uncertainty-Aware Cross-View Consistency, which formulates multi-view projection refinement as a heteroscedastic maximum likelihood estimation problem, utilizing depth confidence as a probabilistic weight. Experiments on dynamic benchmarks show that our approach outperforms current state-of-the-art methods, reducing Mean Accuracy error by 13.43\% and improving segmentation F-measure by 10.49\%. Our framework maintains the efficiency of feed-forward inference and requires no task-specific fine-tuning or per-scene optimization.