GenEAva: Generating Cartoon Avatars with Fine-Grained Facial Expressions from Realistic Diffusion-based Faces

Hao Yu, Rupayan Mallick, Margrit Betke, Sarah Adel Bargal Department of Computer Science, Boston University, Boston, USA Department of Computer Science, Georgetown University, Washington, D.C., USA · 2025

Read GenEAva: Generating Cartoon Avatars with Fine-Grained Facial Expressions from Realistic Diffusion-based Faces by Hao Yu, Rupayan Mallick, Margrit Betke, Sarah Adel Bargal Department of Computer Science, Boston University, Boston, USA Department of Computer Science, Georgetown University, Washington, D.C., USA (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-07945.txt.

Description

Hao Yu and co-authors generate cartoon avatars that pull over a hundred distinct faces. Avatars for games and tutoring need fine expressions, but past sets were bland and often copied real people. Their pipeline first fine tunes an image model to render rich, detailed expressions on synthetic faces. A style model then redraws each face as a cartoon, keeping identity and expression intact.

The result is a new dataset of over thirteen thousand avatars spanning many expressions, balanced across sex, ancestry, and age. Checks show no memorized real identities leak through. The authors offer the set as a benchmark for expressive avatar research.

Catalog record

CREATORHao Yu, Rupayan Mallick, Margrit Betke, Sarah Adel Bargal Department of Computer Science, Boston University, Boston, USA Department of Computer Science, Georgetown University, Washington, D.C., USA
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-07945

Browse the full library · About the Tower of Babel