VMem: Consistent Interactive Video Scene Generation with Surfel-Indexed View Memory

TL;DR

This paper introduces VMem, a surfel-indexed memory module for video scene generation that maintains long-term scene coherence efficiently by selectively recalling relevant past views based on 3D surface geometry.

Contribution

The paper presents a novel surfel-indexed memory module that improves long-term scene coherence and computational efficiency in interactive video scene generation.

Findings

Outperforms existing methods in scene coherence

Enables long-term environment exploration

Reduces computational cost significantly

Abstract

We propose a novel memory module for building video generators capable of interactively exploring environments. Previous approaches have achieved similar results either by out-painting 2D views of a scene while incrementally reconstructing its 3D geometry-which quickly accumulates errors-or by using video generators with a short context window, which struggle to maintain scene coherence over the long term. To address these limitations, we introduce Surfel-Indexed View Memory (VMem), a memory module that remembers past views by indexing them geometrically based on the 3D surface elements (surfels) they have observed. VMem enables efficient retrieval of the most relevant past views when generating new ones. By focusing only on these relevant views, our method produces consistent explorations of imagined environments at a fraction of the computational cost required to use all past views as context. We evaluate our approach on challenging long-term scene synthesis benchmarks and demonstrate superior performance compared to existing methods in maintaining scene coherence and camera control.