Research map / Image generation / Editing and style transfer
MirrorPPR: Exemplar-Based Portrait Photo Retouching
Zhihong Liu, Zheng Li, Jiachun Jin, Siqi Kou, Yitao Jian, Fengpei Yu, Zhijie Deng
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 spatial self-augmentation to self-constructed quadruplets (ablation) | MirrorPPR-Face · SimFace-100 | PSNR: 28.63 -> 32.25 | Improved |
| add spatial self-augmentation to self-constructed quadruplets (ablation) | MirrorPPR-Face · SimFace-100 | Face Similarity: 0.680 -> 0.937 | Improved |
| use self-augmentation instead of cross-identity data (simulated subset) (ablation) | MirrorPPR-Face · SimFace-100 | PSNR: 32.08 -> 32.25 | No clear effect |
| use self-augmentation instead of cross-identity data (professional subset) (ablation) | MirrorPPR-Pro · ProPortrait-500 | PSNR: 30.52 -> 32.65 | Improved |
| use self-augmentation instead of cross-identity data (professional subset) (ablation) | MirrorPPR-Pro · ProPortrait-500 | Face Similarity: 0.916 -> 0.960 | Improved |
| use self-augmentation instead of cross-identity data (ablation) | MirrorPPR · training convergence | training loss | Improved |
| increase learnable query tokens from 8 to 256 (ablation) | MirrorPPR-Face · SimFace-100 | PSNR: 32.25 -> 32.36 | No clear effect |
| increase R-Former layers from 4 to 8 (ablation) | MirrorPPR-Face · SimFace-100 | Face Similarity: 0.937 -> 0.920 | Hurt |
| reduce R-Former layers from 4 to 2 (ablation) | MirrorPPR-Face · SimFace-100 | PSNR: 32.25 -> 32.10 | Hurt |
| increase LoRA rank from 8 to 32 (ablation) | MirrorPPR-Face · SimFace-100 | PSNR: 31.56 -> 32.25 | Improved |
| increase LoRA rank from 32 to 64 (ablation) | MirrorPPR-Face · SimFace-100 | Face Similarity: 0.937 -> 0.933 | Mixed |
| use MirrorPPR instead of best text-guided baseline Nano Banana 2 | MirrorPPR-Pro · ProPortrait-500 | Face Similarity: 0.667 -> 0.960 | Improved |
Closest papers by meaning
- DiffRetouch: Using Diffusion to Retouch on the Shoulder of Experts (AAAI 2025 · similarity 0.86)
- RetouchFormer: Semi-supervised High-Quality Face Retouching Transformer with Prior-Based Selective Self-Attention (AAAI 2024 · similarity 0.85)
- RetouchGPT: LLM-based Interactive High-Fidelity Face Retouching via Imperfection Prompting (AAAI 2025 · similarity 0.85)
- MirrorVerse: Pushing Diffusion Models to Realistically Reflect the World (CVPR 2025 · similarity 0.85)
- InstructPix2NeRF: Instructed 3D Portrait Editing from a Single Image (ICLR 2024 · similarity 0.85)
- PIXELS: Progressive Image Xemplar-based Editing with Latent Surgery (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.
