Reproduce “AvatarMakeup: Realistic Makeup Transfer for 3D Animatable Head Avatars” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Yiming Zhong, Xiaolin Zhang, Ligang Liu, Yao Zhao, Yunchao Wei

A method is presented for transferring realistic makeup from a single reference image onto 3D animatable head avatars. It utilizes a coarse-to-fine strategy with a UV map for consistent multi-view application and a pre-trained diffusion model for detail, achieving a 30.3% improvement in makeup fidelity (DINO-I) and higher realism compared to prior methods.

Key insights

  • Integrating a global UV map to consolidate multi-view outputs from a 2D diffusion model effectively addresses consistency challenges when applying makeup to dynamic 3D avatars.
  • A two-stage 'coarse-to-fine' approach provides a robust mechanism for achieving both structural coherence and intricate detail, balancing dynamic consistency with high-fidelity makeup quality.
  • Image-based conditioning offers superior precision and control for nuanced facial aesthetic tasks compared to language-based prompts in 3D editing applications.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.