NeoVerse: Enhancing 4D World Model with in-the-wild Monocular Videos

TL;DR

NeoVerse introduces a scalable 4D world model that leverages monocular videos for reconstruction and video generation, overcoming previous limitations of data requirements and pre-processing, and achieving state-of-the-art results.

Contribution

NeoVerse presents a novel framework enabling scalable 4D reconstruction and video generation from in-the-wild monocular videos, with pose-free processing and degradation simulation.

Findings

Achieves state-of-the-art performance in 4D reconstruction benchmarks.

Demonstrates versatility across diverse domains.

Operates effectively with monocular videos without extensive pre-processing.

Abstract

In this paper, we propose NeoVerse, a versatile 4D world model that is capable of 4D reconstruction, novel-trajectory video generation, and rich downstream applications. We first identify a common limitation of scalability in current 4D world modeling methods, caused either by expensive and specialized multi-view 4D data or by cumbersome training pre-processing. In contrast, our NeoVerse is built upon a core philosophy that makes the full pipeline scalable to diverse in-the-wild monocular videos. Specifically, NeoVerse features pose-free feed-forward 4D reconstruction, online monocular degradation pattern simulation, and other well-aligned techniques. These designs empower NeoVerse with versatility and generalization to various domains. Meanwhile, NeoVerse achieves state-of-the-art performance in standard reconstruction and generation benchmarks. Our project page is available at https://neoverse-4d.github.io.