Research map / Image generation / Editing and style transfer
Image-to-image translation
Translate images between domains using adversarial and cycle-based networks.
127 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in Editing and style transfer, keyword cluster “translation · style transfer · image image”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Quiet None of the tracked benchmarks has a report from 2025–26.
34 papers in 2023–24 and 16 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 12.5% to 4.1%.
Open questions the data points to
- Papers report SEAN on CelebAMask-HQ (FID) anywhere from 28.1 to 61.06: comparisons against it depend on whose number you take.
- Papers report CoCosNet on DeepFashion (FID) anywhere from 14.4 to 26.9: comparisons against it depend on whose number you take.
No longer reported by papers here: ADE20K (FID, last in 2023); CelebA-HQ (FID, last in 2023); DeepFashion (FID, last in 2023); Horse→Zebra (FID, last in 2024); cat→dog (FID, last in 2024); ADE20K (SWD, 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 · Image-To-Image Translation With Conditional Adversarial Networks (CVPR 2017 · 22,687 citations)
- 2018 · StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translation (CVPR 2018 · 4,243 citations)
- 2019 · Semantic Image Synthesis With Spatially-Adaptive Normalization (CVPR 2019 · 2,923 citations)
- 2020 · StarGAN v2: Diverse Image Synthesis for Multiple Domains (CVPR 2020 · 1,665 citations)
- 2021 · Image-to-Image Translation via Hierarchical Style Disentanglement (CVPR 2021 · 146 citations)
- 2022 · Exploring Patch-Wise Semantic Relation for Contrastive Learning in Image-to-Image Translation Tasks (CVPR 2022 · 112 citations)
- 2023 · BBDM: Image-to-Image Translation With Brownian Bridge Diffusion Models (CVPR 2023 · 251 citations)
- 2024 · StegoGAN: Leveraging Steganography for Non-Bijective Image-to-Image Translation (CVPR 2024 · 59 citations)
- 2025 · Alias-Free Latent Diffusion Models: Improving Fractional Shift Equivariance of Diffusion Latent Space (CVPR 2025 · 6 citations)
- 2026 · Balancing Fidelity and Diversity in Diffusion Models via Symmetric Attention Decomposition: Hopfield Perspective (ICML 2026 · 0 citations)
Earliest papers here
- Image-To-Image Translation With Conditional Adversarial Networks (CVPR 2017 · 22,687 citations)
- Toward Multimodal Image-to-Image Translation (NeurIPS 2017 · 1,021 citations)
- Unsupervised Image-to-Image Translation Networks (NeurIPS 2017 · 508 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.
ADE20K · FID (lower is better)
6 papers, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 24.84 | SEAN: Image Synthesis With Semantic Region-Adaptive Normalization (CVPR 2020 · 350 citations) |
| 2022 | 24.75 | Marginal Contrastive Correspondence for Guided Image Generation (CVPR 2022 · 40 citations) |
CelebA-HQ · FID (lower is better)
6 papers, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 14.3 | Cross-Domain Correspondence Learning for Exemplar-Based Image Translation (CVPR 2020 · 262 citations) |
| 2021 | 12.5 | Unpaired Image-to-Image Translation via Latent Energy Transport (CVPR 2021 · 38 citations) |
| 2023 | 11.5 | Masked and Adaptive Transformer for Exemplar Based Image Translation (CVPR 2023 · 19 citations) |
DeepFashion · FID (lower is better)
6 papers, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 14.4 | Cross-Domain Correspondence Learning for Exemplar-Based Image Translation (CVPR 2020 · 262 citations) |
| 2021 | 13.08 | Unbalanced Feature Transport for Exemplar-Based Image Translation (CVPR 2021 · 63 citations) |
| 2022 | 12.89 | Marginal Contrastive Correspondence for Guided Image Generation (CVPR 2022 · 40 citations) |
| 2023 | 8.2 | Masked and Adaptive Transformer for Exemplar Based Image Translation (CVPR 2023 · 19 citations) |
Horse→Zebra · FID (lower is better)
6 papers, 2019–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2019 | 65.7 | Positional Normalization (NeurIPS 2019 · 8 citations) |
| 2022 | 34.4 | Exploring Patch-Wise Semantic Relation for Contrastive Learning in Image-to-Image Translation Tasks (CVPR 2022 · 112 citations) |
cat→dog · FID (lower is better)
6 papers, 2019–2024 · not reported since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2019 | 47.5 | Positional Normalization (NeurIPS 2019 · 8 citations) |
| 2021 | 45.2 | Unpaired Image-to-Image Translation via Latent Energy Transport (CVPR 2021 · 38 citations) |
ADE20K · SWD (lower is better)
5 papers, 2020–2023 · not reported since 2023
| Year | Best so far | Set by |
|---|---|---|
| 2020 | 10.5 | Cross-Domain Correspondence Learning for Exemplar-Based Image Translation (CVPR 2020 · 262 citations) |
| 2021 | 9.9 | CoCosNet v2: Full-Resolution Correspondence Learning for Image Translation (CVPR 2021 · 118 citations) |
| 2022 | 9.852 | Marginal Contrastive Correspondence for Guided Image Generation (CVPR 2022 · 40 citations) |
What has been tried
112 papers here report 1,173 measured changes (a component added, removed or swapped, compared with a baseline): helped 72%, no clear effect 5%, hurt 20%, 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 |
|---|---|---|---|---|
| vary encoder depth from the full method's depth 1 | PlanIGN · RMSE | 22.5 -> 22.4 (depth -1); 22.5 -> 22.8 (depth 3); 22.5 -> 24.2 (depth 5); 22.5 -> 24.5 (depth 8) | Mixed | StegoGAN: Leveraging Steganography for Non-Bijective Image-to-Image Translation (CVPR 2024 · 59 citations) |
| Add Prior Mixup | LibriTTS many-to-many · EER | 7.99 -> 4.24 | Helped | DDDM-VC: Decoupled Denoising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust Voice Conversion (AAAI 2024 · 35 citations) |
| Add a second same-stage feature pair (h13,h18) | Cat→Dog · FID | 63.8 -> 63.8 | No clear effect | Rethinking the Paradigm of Content Constraints in Unpaired Image-to-Image Translation (AAAI 2024 · 18 citations) |
| Add DRF | Places2, all reported mask ratios · PSNR | 23.3944 -> 24.0885; 21.7255 -> 22.3871; 19.3818 -> 19.9352 | Helped | SyFormer: Structure-Guided Synergism Transformer for Large-Portion Image Inpainting (AAAI 2024 · 16 citations) |
| add perpendicular regularization | Sketches, Babies, Sunglasses, Amedeo, Raphael, Otto, Haunted, Wrecked · FID/KID | Mixed | WeditGAN: Few-Shot Image Generation via Latent Space Relocation (AAAI 2024 · 13 citations) | |
| Add Equifrequency Shuffle to the mini-frequency branch | Style-guided content creation · spatial structure disentanglement | Helped | Frequency-Controlled Diffusion Model for Versatile Text-Guided Image-to-Image Translation (AAAI 2024 · 10 citations) | |
| apply instance normalization guidance instead of classifier-free guidance with w = 2.5 | Instance normalization study · CLIP-Score | 0.833 -> 0.839 | Helped | Image Translation as Diffusion Visual Programmers (ICLR 2024 · 10 citations) |
| Add SegRefine to the initial segmentation pipeline | AASCE2019 test · MAE_avg | 4.64 -> 4.22 | Helped | B-spine: Learning B-spline Curve Representation for Robust and Interpretable Spinal Curvature Estimation (AAAI 2024 · 9 citations) |
| add a patch discriminator instead of an instance discriminator | Horse2Zebra · FID | 104 -> 66.3 | Helped | Unpaired Image-to-Image Translation via Neural Schrödinger Bridge (ICLR 2024 · 8 citations) |
| Add FDA to the single-scale architecture | SYNTHIA Day →Night · SSIM | 0.54 -> 0.67 | Helped | SEIT: Structural Enhancement for Unsupervised Image Translation in Frequency Domain (AAAI 2024 · 7 citations) |
| Add ideal sampling | ImageNet 256×256 validation · encoder SPSNR | 19.71 -> 21.51 | Helped | Alias-Free Latent Diffusion Models: Improving Fractional Shift Equivariance of Diffusion Latent Space (CVPR 2025 · 6 citations) |
| Set CFG scale to 8.0 | Boson-night · FID | 161.22 -> 116.46 | Helped | ThermalGen: Style-Disentangled Flow-Based Generative Models for RGB-to-Thermal Image Translation (NeurIPS 2025 · 4 citations) |
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 |
|---|---|---|---|---|
| SEAN | CelebAMask-HQ | FID | 28.1 – 61.06 | 2 papers, e.g. NeurIPS 2023 and AAAI 2021 |
| CoCosNet | DeepFashion | FID | 14.4 – 26.9 | 2 papers, e.g. CVPR 2021 and CVPR 2021 |
| CoCosNet | DeepFashion | SWD | 17.2 – 29 | 2 papers, e.g. CVPR 2021 and CVPR 2021 |
| CoCosNet v2 | Deepfashion | SWD | 16.53 – 24.6 | 2 papers, e.g. CVPR 2022 and AAAI 2023 |
| CoCosNet v2 | CelebA-HQ | SWD | 14.62 – 19.7 | 2 papers, e.g. CVPR 2022 and AAAI 2023 |
| CoCosNet v2 | CelebA-HQ | FID | 12.85 – 16.2 | 2 papers, e.g. CVPR 2022 and AAAI 2023 |
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.
Baselines
- Pix2Pix · 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 50 papers; from OpenAlex and the papers' affiliations).
- Chinese Academy of Sciences · 2 papers
- Korea University · 2 papers
- Yonsei University · 2 papers
- Peking University · 2 papers
- Xi’an Jiaotong University · 2 papers
- Zhejiang University · 2 papers
- Shanghai Jiao Tong University · 2 papers
- Korea Advanced Institute of Science and Technology · 2 papers
- Xidian University · 2 papers
- Southeast 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
- Texture and sketch synthesis (130 papers)
- Arbitrary neural style transfer (124 papers)
- Virtual try-on (46 papers)
- Pose-guided person image synthesis (29 papers)
Elsewhere in the topic
- Instruction-based diffusion editing (151 papers)
- Inversion-based diffusion editing (113 papers)
- Identity-preserving face manipulation (88 papers)
- Localized and attribute-specific editing (73 papers)
- Diffusion face generation and editing (61 papers)
- GAN inversion for editing (25 papers)
- Talking face generation (21 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).