LiveScene: Language Embedding Interactive Radiance Fields for Physical Scene Rendering and Control

TL;DR

LiveScene introduces a novel scene-level radiance field model that uses language embeddings for interactive scene reconstruction and control, enabling efficient, accurate, and natural language-guided manipulation of complex scenes.

Contribution

The paper presents LiveScene, a new method that decomposes scenes into local deformable fields and uses language embeddings for interactive control, advancing scene reconstruction and manipulation.

Findings

Outperforms existing methods in novel view synthesis

Enables natural language-based scene control

Reduces memory consumption in scene reconstruction

Abstract

This paper scales object-level reconstruction to complex scenes, advancing interactive scene reconstruction. We introduce two datasets, OmniSim and InterReal, featuring 28 scenes with multiple interactive objects. To tackle the challenge of inaccurate interactive motion recovery in complex scenes, we propose LiveScene, a scene-level language-embedded interactive radiance field that efficiently reconstructs and controls multiple objects. By decomposing the interactive scene into local deformable fields, LiveScene enables separate reconstruction of individual object motions, reducing memory consumption. Additionally, our interaction-aware language embedding localizes individual interactive objects, allowing for arbitrary control using natural language. Our approach demonstrates significant superiority in novel view synthesis, interactive scene control, and language grounding performance through extensive experiments. Project page: https://livescenes.github.io.