Research map / Image generation / Editing and style transfer
Anchoring on Reality: Breaking the Pseudo-Target Ceiling in Makeup Transfer
Bo Wei, Xianhui Lin, Yi Dong, Zhongzhong Li, Zonghui Li, Zirui Wang, Jiachen Yang, Xing Liu, Hong Gu, Xiaoming Li, Wangmeng Zuo
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Stage II reality-anchored refinement to Stage I (ablation) | ART · MF2K test set | MSim_G: 8.82 -> 9.22 | Improved |
| add Stage II refinement to Stage I (ablation) | ART · MF2K test set | ID similarity: .57 -> .74 | Improved |
| add Stage I initialization to Stage II (ablation) | ART · MF2K test set | MSim_G: 8.08 -> 9.22 | Improved |
| add Stage I initialization to Stage II (ablation) | ART · MT test set | MSim_Q: 5.35 -> 7.12 | Hurt |
| add Stage II refinement to Stage I (ablation) | ART · MF2K test set | L2-M: 7.68 -> 4.31 | Improved |
| add Stage II refinement to Stage I (ablation) | ART · MT test set | FID: 84.93 -> 87.72 | Hurt |
| train on StableMakeup pseudo-targets instead of Nano Banana Pro (ablation) | ART · MF2K test set | MSim_G: 9.22 -> 8.56 | Hurt |
| apply ART refinement on StableMakeup pseudo-targets vs StableMakeup (ablation) | ART (SM pseudo-targets) · MF2K test set | MSim_G: 6.73 -> 8.56 | Improved |
| retrain StableMakeup on MF2K data (ablation) | StableMakeup · MF2K test set | MSim_G: 6.73 -> 6.58 | No clear effect |
| retrain StableMakeup on MF2K data (ablation) | StableMakeup · MT test set | MSim_G: 8.34 -> 8.47 | Improved |
| add ~4,500 unpaired images to Stage II refinement (ablation) | ART · MF2K test set | MSim_G: 9.14 -> 9.22 | Improved |
| increase residual makeup alpha in bare-skin anchor from 0.00 to 0.10 (ablation) | ART · Four test sets (average) | MSim_G: 9.02 -> 8.84 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PSGAN — PSGAN: Pose and Expression Robust Spatial-Aware GAN for Customizable Makeup Transfer (CVPR 2020) · better on 1 of 1 shared result
- EleGANt — ELEGANT: Exchanging Latent Encodings with GAN for Transferring Multiple Face Attributes (ECCV 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SHMT: Self-supervised Hierarchical Makeup Transfer via Latent Diffusion Models (NeurIPS 2024 · similarity 0.89)
- Makeup Prior Models for 3D Facial Makeup Estimation and Applications (CVPR 2024 · similarity 0.87)
- Content-Style Decoupling for Unsupervised Makeup Transfer without Generating Pseudo Ground Truth (CVPR 2024 · similarity 0.87)
- DreamSalon: A Staged Diffusion Framework for Preserving Identity-Context in Editable Face Generation (CVPR 2024 · similarity 0.86)
- When StyleGAN Meets Stable Diffusion: a W+ Adapter for Personalized Image Generation (CVPR 2024 · similarity 0.85)
- What to Preserve and What to Transfer: Faithful, Identity-Preserving Diffusion-based Hairstyle Transfer (AAAI 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
