4DGS-Fixer: Video Diffusion Priors for Sparse-View 4D Scenes
A new method iteratively refines 4D Gaussian splatting with video diffusion priors to improve dynamic scene synthesis from sparse-view videos.
Reconstructing a dynamic scene from only a few camera viewpoints is a difficult problem, because large parts of the scene are never directly seen. The new paper introduces 4DGS-Fixer, a method that builds on 4D Gaussian Splatting and incorporates generative video diffusion priors into the reconstruction process.
The authors note that prior work typically handles sparse-view dynamic scenes with geometric priors, adaptive optimization, or density-control strategies. 4DGS-Fixer is positioned around an iterative refinement loop in which video diffusion priors guide the 4D Gaussian model, offering a generative complement to those more traditional constraints.
The significance is in the approach: rather than treating sparse views as a purely geometric or optimization problem, the method brings learned video priors directly into the refinement process. This points toward a more robust route for synthesizing dynamic scenes when input observations are limited.
Sources · 10
- Splat-CBF: Safe Next-Best-View Control in 3D Gaussian-Splat Maps
- D3GS: Depth, DINO, and RGB Diffusion Co-Guided 3D Gaussian Splatting for Sparse-View Reconstruction
- Compressing 3D Gaussian Splatting via Cross-Representation Priors
- GrapeSplat: Geometry-Grounded Reconstruction via Amalgamated Pose-Free Encoding for Feed-Forward 3D Gaussian Splatting
- VDGS: Visibility-Driven Large-Scale 3D Gaussian Splatting for Aerial Scene Reconstruction
- LiteTex-GS: Fast and Lightweight Texturing for Gaussian Splatting
- GAPS: Generative Active Pseudo-view Selection for Sparse-View 3D Gaussian Splatting
- OpenFlyScan: A Quality-Guided Aerial Reconstruction System for Consumer Drones
- GARO: Geometry-Aware Redundancy Optimization for Real-Time and High-Fidelity Dynamic Gaussian Splatting
- 4DGS-Fixer: Generative Sparse-View 4D Gaussian Splatting with Iterative Refinement Guided by Video Diffusion Priors
More in Research Digest
OpenRoIS middleware decouples HRI apps from robot hardware
A new open-source middleware aims to let human-robot interaction applications run across different robot platforms without rewriting code.
Two papers probe RL bottlenecks for LLM agents
New arXiv work targets two stubborn obstacles in reinforcement learning for LLMs: sparse credit in open-ended tasks and merging separately trained capabilities.
LLM Social Agents Need More Than Mimicry, Two Preprints Argue
Two new arXiv papers identify a common reliability gap in LLM-driven social simulations, then propose different fixes: tuned personalities versus a Bayesian belief layer.
Proactive Robots Must Balance Helpfulness Against Intrusiveness
Three new papers examine how robots can anticipate human actions, time their assistance, and tailor it to individual users without becoming intrusive.