Anchoring on Reality: Breaking the Pseudo-Target Ceiling in Makeup Transfer

2026-06-29

AI Summary

This work addresses the challenges of detail degradation, synthesis artifacts, and identity drift in makeup transfer, which stem from the scarcity of real paired data. To overcome these issues, the authors propose ART, a two-stage framework: the first stage leverages large-model-generated pseudo-targets to achieve semantic alignment, while the second introduces a novel reality-anchored differentiable cycle mechanism that reconstructs makeup images from bare faces under the supervision of real reference images. To facilitate research in this domain, the authors also construct MakeupFaces2K (MF2K), the first in-the-wild makeup portrait dataset at 2K resolution. Experimental results demonstrate that the proposed method significantly outperforms existing approaches in makeup fidelity, background stability, and identity preservation, particularly under complex makeup conditions.

Abstract

Makeup transfer applies a reference cosmetic style to a source face while preserving its identity and geometry. However, this task is severely hindered by the lack of real paired training data. Current methods rely on either weak priors or synthetic pseudo-targets from large-scale editing models. These paradigms provide suboptimal guidance, often leading to degraded fine-grained details, synthetic artifacts, and identity drift. To this end, we propose Anchoring on Reality Makeup Transfer (ART), a two-stage framework with a reality-anchored refinement cycle. In Stage I, the model is initialized with pseudo-targets to establish basic semantic alignment and global makeup placement. Crucially, Stage II shifts supervision from pseudo-targets to the real reference, reconstructing it from its bare-skin counterpart through a differentiable cycle that penalizes any omitted detail and overrides synthetic artifacts. Furthermore, we introduce MakeupFaces2K (MF2K), the first 2K-resolution in-the-wild makeup portrait dataset comprising 8,573 images. Extensive experiments demonstrate that our method achieves superior makeup fidelity, strong background stability, and robust identity preservation, especially for complex makeup styles.

Problem

• makeup transfer
• paired training data
• pseudo-targets
• identity preservation
• synthetic artifacts

Innovation

• makeup transfer
• reality-anchored refinement
• pseudo-target
• differentiable cycle
• high-resolution dataset