DreamCube: 3D Panorama Generation via Multi-plane Synchronization

TL;DR

DreamCube introduces a novel multi-plane synchronization method that extends 2D foundation models to generate high-quality, diverse 3D panoramas with consistent geometry and appearance.

Contribution

The paper presents DreamCube, a multi-plane RGB-D diffusion model that effectively adapts 2D foundation models for 3D panorama synthesis, addressing previous limitations.

Findings

Effective panoramic image generation

Accurate panoramic depth estimation

Consistent 3D scene generation

Abstract

3D panorama synthesis is a promising yet challenging task that demands high-quality and diverse visual appearance and geometry of the generated omnidirectional content. Existing methods leverage rich image priors from pre-trained 2D foundation models to circumvent the scarcity of 3D panoramic data, but the incompatibility between 3D panoramas and 2D single views limits their effectiveness. In this work, we demonstrate that by applying multi-plane synchronization to the operators from 2D foundation models, their capabilities can be seamlessly extended to the omnidirectional domain. Based on this design, we further introduce DreamCube, a multi-plane RGB-D diffusion model for 3D panorama generation, which maximizes the reuse of 2D foundation model priors to achieve diverse appearances and accurate geometry while maintaining multi-view consistency. Extensive experiments demonstrate the effectiveness of our approach in panoramic image generation, panoramic depth estimation, and 3D scene generation.