4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos

Zhen Xu (Institute of Information Engineering, CAS) · Zhengqin Li (Meta Reality Labs) · Zhao Dong (Meta RL Research) · Xiaowei Zhou (Zhejiang University) · Richard Newcombe (Meta, Reality Labs Research) · Zhaoyang Lv

4d gaussiancross-domain videosdensity control strategydynamic scene reconstruction feed-forward inferencegaussian-based networksinductive bias large-scale datasetsmonocular posed videosobject lifespansreconstruction timerolling-window fashionspace-time inputtime-varying environmentstransformer model

We propose 4DGT, a 4D Gaussian-based Transformer model for dynamic scene reconstruction, trained entirely on real-world monocular posed videos. Using 4D Gaussian as an inductive bias, 4DGT unifies static and dynamic components, enabling the modeling of complex, time-varying environments with varying object lifespans. We proposed a novel density control strategy in training, which enables our 4DGT to handle longer space-time input. Our model processes 64 consecutive posed frames in a rolling-window fashion, predicting consistent 4D Gaussians in the scene. Unlike optimization-based methods, 4DGT performs purely feed-forward inference, reducing reconstruction time from hours to seconds and scaling effectively to long video sequences. Trained only on large-scale monocular posed video datasets, 4DGT can outperform prior Gaussian-based networks significantly in real-world videos and achieve on-par accuracy with optimization-based methods on cross-domain videos.