Saved in:
Bibliographic Details
Main Authors: Gridusov, Denis, Popov, Maxim, Kolyubin, Sergey
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.25909
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910255351332864
author Gridusov, Denis
Popov, Maxim
Kolyubin, Sergey
author_facet Gridusov, Denis
Popov, Maxim
Kolyubin, Sergey
contents Reconstructing and predicting dynamic 3D scenes from multi-view videos is a foundational task for robotics, AR/VR, and digital twins. Recent physics-informed Gaussian Splatting methods achieve impressive future frame extrapolation but lack semantic awareness and suffer from large computational overhead. We introduce $\textbf{R5DGS}$, a framework that augments a physics-driven 4D Gaussian representation with compact Identity Encoding vectors, enabling precise Gaussian-to-object association. By constructing an offline CLIP-based object lookup table, we support open-vocabulary text prompting to retrieve and render object-specific Gaussians across arbitrary timestamps and viewpoints. Furthermore, we propose a rigid-body inference constraint that predicts and integrates physical dynamics exclusively for object centroids, propagating motion to associated Gaussians via relative transformations. This optimization yields a 11 FPS speedup during extrapolation without compromising trajectories plausibility.
format Preprint
id arxiv_https___arxiv_org_abs_2605_25909
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle R5DGS: Semantic-Aware 4D Gaussian Splatting with Rigid Body Constraints for Efficient Dynamic Scene Reconstruction
Gridusov, Denis
Popov, Maxim
Kolyubin, Sergey
Computer Vision and Pattern Recognition
Reconstructing and predicting dynamic 3D scenes from multi-view videos is a foundational task for robotics, AR/VR, and digital twins. Recent physics-informed Gaussian Splatting methods achieve impressive future frame extrapolation but lack semantic awareness and suffer from large computational overhead. We introduce $\textbf{R5DGS}$, a framework that augments a physics-driven 4D Gaussian representation with compact Identity Encoding vectors, enabling precise Gaussian-to-object association. By constructing an offline CLIP-based object lookup table, we support open-vocabulary text prompting to retrieve and render object-specific Gaussians across arbitrary timestamps and viewpoints. Furthermore, we propose a rigid-body inference constraint that predicts and integrates physical dynamics exclusively for object centroids, propagating motion to associated Gaussians via relative transformations. This optimization yields a 11 FPS speedup during extrapolation without compromising trajectories plausibility.
title R5DGS: Semantic-Aware 4D Gaussian Splatting with Rigid Body Constraints for Efficient Dynamic Scene Reconstruction
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.25909