WildAvatar: Learning In-the-wild 3D Avatars from the Web

TL;DR

WildAvatar introduces a scalable pipeline for creating diverse 3D human avatars from web videos, resulting in a large, real-world dataset that enhances avatar creation methods and broadens practical applications.

Contribution

The paper presents an automatic annotation pipeline and a new large-scale in-the-wild dataset, WildAvatar, significantly surpassing previous datasets in size and real-world diversity for 3D avatar creation.

Findings

Pipeline outperforms state-of-the-art on EMDB benchmark

Filtering protocols improve web video verification metrics

WildAvatar dataset is 10 times richer than previous datasets

Abstract

Existing research on avatar creation is typically limited to laboratory datasets, which require high costs against scalability and exhibit insufficient representation of the real world. On the other hand, the web abounds with off-the-shelf real-world human videos, but these videos vary in quality and require accurate annotations for avatar creation. To this end, we propose an automatic annotating pipeline with filtering protocols to curate these humans from the web. Our pipeline surpasses state-of-the-art methods on the EMDB benchmark, and the filtering protocols boost verification metrics on web videos. We then curate WildAvatar, a web-scale in-the-wild human avatar creation dataset extracted from YouTube, with $10000+$ different human subjects and scenes. WildAvatar is at least $10\times$ richer than previous datasets for 3D human avatar creation and closer to the real world. To explore its potential, we demonstrate the quality and generalizability of avatar creation methods on WildAvatar. We will publicly release our code, data source links and annotations to push forward 3D human avatar creation and other related fields for real-world applications.