arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2307.05445cs.CV

自动解码潜空间3D扩散模型

AutoDecoding Latent 3D Diffusion Models

ETH Zurich(苏黎世联邦理工学院);Snap Inc.(Snap公司);Carnegie Mellon University(卡内基梅隆大学);KU Leuven(鲁汶大学)

Evangelos Ntavelis, Aliaksandr Siarohin, Kyle Olszewski, Chaoyang Wang, Luc Van Gool, Sergey Tulyakov

AI总结:

本文提出以3D自动解码器为核心的潜空间3D扩散模型,利用2D图像或单目视频学习体积潜表示,并支持有相机监督或无相机训练,在多类3D生成基准上超过现有方法。

摘要:

我们提出了一种生成静态和铰接式3D资产的新颖方法,其核心是一个3D自动解码器。该3D自动解码器框架将从目标数据集学到的属性嵌入到潜空间中,随后可将其解码为体积表示,以渲染视角一致的外观和几何。接着,我们确定了合适的中间体积潜空间,并引入鲁棒的归一化与反归一化操作,从而从刚性或铰接对象的2D图像或单目视频中学习3D扩散。我们的方法足够灵活,既可以使用现有的相机监督,也可以完全不使用相机信息——转而在训练过程中高效地学习相机信息。评估表明,在多种基准数据集和指标上,我们的生成结果优于最先进的替代方法,包括合成对象的多视图图像数据集、移动人物的真实野外视频,以及一个大规模静态对象真实视频数据集。

英文摘要:

We present a novel approach to the generation of static and articulated 3D assets that has a 3D autodecoder at its core. The 3D autodecoder framework embeds properties learned from the target dataset in the latent space, which can then be decoded into a volumetric representation for rendering view-consistent appearance and geometry. We then identify the appropriate intermediate volumetric latent space, and introduce robust normalization and de-normalization operations to learn a 3D diffusion from 2D images or monocular videos of rigid or articulated objects. Our approach is flexible enough to use either existing camera supervision or no camera information at all -- instead efficiently learning it during training. Our evaluations demonstrate that our generation results outperform state-of-the-art alternatives on various benchmark datasets and metrics, including multi-view image datasets of synthetic objects, real in-the-wild videos of moving people, and a large-scale, real video dataset of static objects.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑