Research map / Image generation / Editing and style transfer
Identity-preserving face manipulation
Edit, swap, and reenact faces with GAN-based disentanglement while keeping identity.
88 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Editing and style transfer, keyword cluster “face · facial · identity”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Quiet None of the tracked benchmarks has a report from 2025–26.
22 papers in 2023–24 and 19 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 8.1% to 4.8%.
Open questions the data points to
- Papers report AttnGAN on Multi-modal CelebA-HQ (FID) anywhere from 54.22 to 125.98: comparisons against it depend on whose number you take.
- Papers report MegaFS on CelebA-HQ (FID) anywhere from 11.64 to 18.81: comparisons against it depend on whose number you take.
No longer reported by papers here: CelebA-HQ (FID, last in 2023).
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Generative Face Completion (CVPR 2017 · 619 citations)
- 2018 · Facial Expression Recognition by De-Expression Residue Learning (CVPR 2018 · 447 citations)
- 2019 · Mask-Guided Portrait Editing With Conditional GANs (CVPR 2019 · 143 citations)
- 2020 · Face X-Ray for More General Face Forgery Detection (CVPR 2020 · 1,140 citations)
- 2021 · TediGAN: Text-Guided Diverse Face Image Generation and Manipulation (CVPR 2021 · 360 citations)
- 2022 · Protecting Celebrities From DeepFake With Identity Consistency Transformer (CVPR 2022 · 170 citations)
- 2023 · Fine-Grained Face Swapping via Regional GAN Inversion (CVPR 2023 · 85 citations)
- 2024 · Generative Unlearning for Any Identity (CVPR 2024 · 14 citations)
- 2025 · SlerpFace: Face Template Protection via Spherical Linear Interpolation (AAAI 2025 · 10 citations)
- 2026 · Optimizing ID Consistency in Multimodal Large Models: Facial Restoration via Alignment, Entanglement, and Disentanglement (ICLR 2026 · 0 citations)
Earliest papers here
- Generative Face Completion (CVPR 2017 · 619 citations)
- Neural Face Editing With Intrinsic Image Disentangling (CVPR 2017 · 307 citations)
- Learning Residual Images for Face Attribute Manipulation (CVPR 2017 · 289 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
CelebA-HQ · FID (lower is better)
6 papers, 2021–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 4.04 | High-Fidelity and Arbitrary Face Editing (CVPR 2021 · 88 citations) |
What has been tried
77 papers here report 708 measured changes (a component added, removed or swapped, compared with a baseline): helped 74%, no clear effect 4%, hurt 19%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use an extrapolated target latent instead of the average latent | Random, In-Domain (FFHQ), and Out-of-Domain (CelebAHQ) · ID | 0.19 ± 0.09 -> 0.12 ± 0.06; 0.16 ± 0.07 -> 0.05 ± 0.06; 0.12 ± 0.06 -> 0.02 ± 0.05 | Helped | Generative Unlearning for Any Identity (CVPR 2024 · 14 citations) |
| seamless mix blending (vs. linear blending) on neural warping | FRLL morphings, t=0.5 · LPIPS (I0, I) | 0.158 -> 0.081 | Helped | Neural Implicit Morphing of Face Images (CVPR 2024 · 13 citations) |
| add learnable feature grouping and weighted dropout | CFP-FP · accuracy | 87.13 -> 92.79 | Helped | SlerpFace: Face Template Protection via Spherical Linear Interpolation (AAAI 2025 · 10 citations) |
| Add negative identifiers to positive identifiers | One-shot face stylization · Content preservation | Helped | StyO: Stylize Your Face in Only One-Shot (AAAI 2025 · 5 citations) | |
| Add Align-CA to the baseline | K-Hairstyle · SSIM | 0.58 -> 0.62 | Helped | What to Preserve and What to Transfer: Faithful, Identity-Preserving Diffusion-based Hairstyle Transfer (AAAI 2025 · 4 citations) |
| add multi-scale tokens rather than using a single patch size | FFHQR (γ=0.1) · PSNR | 40.20 -> 45.13 | Helped | RetouchFormer: Semi-supervised High-Quality Face Retouching Transformer with Prior-Based Selective Self-Attention (AAAI 2024 · 4 citations) |
| use grayscale input and Sobel gradients instead of color input and Sobel gradients | CariFace · NME | 7.863 -> 7.695 | Helped | Generalizable Face Landmarking Guided by Conditional Face Warping (CVPR 2024 · 4 citations) |
| Add Steps 2+3 after Step 1 | CelebA-HQ · Recall@k=1 | 8.08 -> 0.015 | Helped | ToonerGAN: Reinforcing GANs for Obfuscating Automated Facial Indexing (CVPR 2024 · 4 citations) |
| Add CMM and RCG to FaceNet | FFHQ · DECA re-inference error | 7.13 -> 4.85 | Helped | ControlFace: Harnessing Facial Parametric Control for Face Rigging (CVPR 2025 · 3 citations) |
| add ID patch projection | Evaluation set · ID-position association accuracy | 0.345 -> 0.958 | Helped | ID-Patch: Robust ID Association for Group Photo Personalization (CVPR 2025 · 2 citations) |
| Add the hair injection module | Head swapping · hairstyle detail | Helped | Zero-Shot Head Swapping in Real-World Scenarios (CVPR 2025 · 2 citations) | |
| Add FASA without APG | Personalization models grouped by foundation model · PC (%) | SDv1.5 21.83; SDXL 24.55; FLUX 24.50 | Helped | Foundation Cures Personalization: Improving Personalized Models’ Prompt Consistency via Hidden Foundation Knowledge (NeurIPS 2025 · 1 citation) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| AttnGAN | Multi-modal CelebA-HQ | FID | 54.22 – 125.98 | 2 papers, e.g. CVPR 2022 and CVPR 2021 |
| MegaFS | CelebA-HQ | FID | 11.64 – 18.81 | 2 papers, e.g. CVPR 2022 and CVPR 2022 |
| ControlGAN | Multi-modal CelebA-HQ | FID | 77.41 – 116.32 | 2 papers, e.g. CVPR 2022 and CVPR 2021 |
| ControlGAN | Multi-modal CelebA-HQ | LPIPS | 0.522 – 0.572 | 2 papers, e.g. CVPR 2021 and CVPR 2022 |
| AttnGAN | Multi-modal CelebA-HQ | LPIPS | 0.512 – 0.548 | 2 papers, e.g. CVPR 2021 and CVPR 2022 |
What reviewers will expect
Datasets and baselines named in the 10 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- MS1Mv2 · 2 of 10 papers
- LFW · 2 of 10 papers
- CFP-FP · 2 of 10 papers
- CPLFW · 2 of 10 papers
- CALFW · 2 of 10 papers
- IJB-B · 2 of 10 papers
- FFHQ · 2 of 10 papers
Baselines
- Arc2Face · 3 of 10 papers
- IDiff-Face · 2 of 10 papers
- DCFace · 2 of 10 papers
- ID^3 · 2 of 10 papers
- DigiFace · 2 of 10 papers
- SynFace · 2 of 10 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 41 papers; from OpenAlex and the papers' affiliations).
- South China University of Technology · 4 papers
- City University of Hong Kong · 3 papers
- University of Science and Technology of China · 2 papers
- Shantou University · 2 papers
- Tencent · 2 papers
- Korea Advanced Institute of Science and Technology · 2 papers
- Pazhou Lab · 2 papers
- Peng Cheng Laboratory · 2 papers
- Shanghai Jiao Tong University · 2 papers
Use it
Download BibTeX of 40 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Diffusion face generation and editing (61 papers)
- Talking face generation (21 papers)
Elsewhere in the topic
- Instruction-based diffusion editing (151 papers)
- Texture and sketch synthesis (130 papers)
- Image-to-image translation (127 papers)
- Arbitrary neural style transfer (124 papers)
- Inversion-based diffusion editing (113 papers)
- Localized and attribute-specific editing (73 papers)
- Virtual try-on (46 papers)
- Pose-guided person image synthesis (29 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).