Research map / Image generation / Editing and style transfer
Arbitrary neural style transfer
Transfer artistic style with attention, feature transforms, and brushstroke models.
124 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Editing and style transfer, keyword cluster “translation · style transfer · image image”
State History Tried Baselines Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
24 papers in 2023–24 and 33 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic held steady, from 8.8% to 8.4%.
Open questions the data points to
- Papers report AdaIN on 256×256 (execution time (sec)) anywhere from 0.053 to 0.09: comparisons against it depend on whose number you take.
- Papers report AdaIN on 512×512 (execution time (sec)) anywhere from 0.11 to 0.17: comparisons against it depend on whose number you take.
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
- 2016 · Image Style Transfer Using Convolutional Neural Networks (CVPR 2016 · 6,164 citations)
- 2017 · A Learned Representation For Artistic Style (ICLR 2017 · 876 citations)
- 2018 · PairedCycleGAN: Asymmetric Style Transfer for Applying and Removing Makeup (CVPR 2018 · 321 citations)
- 2019 · Arbitrary Style Transfer With Style-Attentional Networks (CVPR 2019 · 476 citations)
- 2020 · Dynamic Instance Normalization for Arbitrary Style Transfer (AAAI 2020 · 197 citations)
- 2021 · Encoding in Style: A StyleGAN Encoder for Image-to-Image Translation (CVPR 2021 · 1,083 citations)
- 2022 · CLIPstyler: Image Style Transfer With a Single Text Condition (CVPR 2022 · 246 citations)
- 2023 · Inversion-Based Style Transfer With Diffusion Models (CVPR 2023 · 345 citations)
- 2024 · Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 2024 · 202 citations)
- 2025 · Attention Distillation: A Unified Approach to Visual Characteristics Transfer (CVPR 2025 · 28 citations)
- 2026 · An Efficient and Harmonized Framework for Balanced Cross-Domain Feature Integration (AAAI 2026 · 2 citations)
Earliest papers here
- Image Style Transfer Using Convolutional Neural Networks (CVPR 2016 · 6,164 citations)
- Texture Networks: Feed-forward Synthesis of Textures and Stylized Images (ICML 2016 · 599 citations)
- Split and Match: Example-Based Adaptive Patch Sampling for Unsupervised Style Transfer (CVPR 2016 · 132 citations)
What has been tried
105 papers here report 935 measured changes (a component added, removed or swapped, compared with a baseline): helped 71%, no clear effect 5%, hurt 21%, 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 |
|---|---|---|---|---|
| remove style injection | Style transfer · ArtFID | 28.80 -> 43.72 | Hurt | Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 2024 · 202 citations) |
| add the text prompt | Van Gogh, Morisot, Ukiyoe, Monet, Cezanne, Gauguin, Peploe · CLIP score | Van Gogh: 0.7248 -> 0.7321 | Helped | ArtBank: Artistic Style Transfer with Pre-trained Diffusion Model and Implicit Style Prompt Bank (AAAI 2024 · 63 citations) |
| Start attention injection later in denoising | Style-transfer examples · content preservation | Helped | Z*: Zero-shot Style Transfer via Attention Reweighting (CVPR 2024 · 52 citations) | |
| remove StyleRemover from the content encoder | HumanML3D test set · SRA (%) | 16.88 -> 58.00 | Hurt | Arbitrary Motion Style Transfer with Multi-condition Motion Latent Diffusion Model (CVPR 2024 · 30 citations) |
| Add the Auxiliary Content Adapter during training | Style transfer · Content separation | Helped | ArtAdapter: Text-to-Image Style Transfer using Multi-Level Style Encoder and Explicit Adaptation (CVPR 2024 · 29 citations) | |
| Use the Adam optimizer to manage guidance strength instead of manually setting it | style and appearance transfer · visual match to the reference | Helped | Attention Distillation: A Unified Approach to Visual Characteristics Transfer (CVPR 2025 · 28 citations) | |
| Use content and style feature extractors before the encoder-based transformer | Style transfer · visual content preservation and stylization | Helped | Puff-Net: Efficient Style Transfer with Pure Content and Style Feature Fusion Network (CVPR 2024 · 24 citations) | |
| Remove Top-K selection and use fixed timestep-based selection | SD and FLUX object-style fusion · Visual preservation of subject and style | Hurt | K-LoRA: Unlocking Training-Free Fusion of Any Subject and Style LoRAs (CVPR 2025 · 20 citations) | |
| add aesthetic feature contrastive loss | Stylization results (ablation study) · Stylization quality | Helped | AesFA: An Aesthetic Feature-Aware Arbitrary Neural Style Transfer (AAAI 2024 · 19 citations) | |
| replace PGT-guided training with the unsupervised CSD-MT strategy | MT · Self-Aug PSNR (Crop) | 23.23 -> 27.28 | Helped | Content-Style Decoupling for Unsupervised Makeup Transfer without Generating Pseudo Ground Truth (CVPR 2024 · 19 citations) |
| Add zigzag scan instead of cross scan | Ablation study · ArtFID | 26.808 -> 26.305 | Helped | SaMam: Style-aware State Space Model for Arbitrary Image Style Transfer (CVPR 2025 · 17 citations) |
| Add Cross-Modal AdaIN and the Teacher Model | 52 prompts and 20 style reference images · text alignment | 0.216 -> 0.235 | Helped | StyleStudio: Text-Driven Style Transfer with Selective Control of Style Elements (CVPR 2025 · 17 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 |
|---|---|---|---|---|
| AdaIN | 256×256 | execution time (sec) | 0.053 – 0.09 | 2 papers, e.g. CVPR 2018 and CVPR 2019 |
| AdaIN | 512×512 | execution time (sec) | 0.11 – 0.17 | 2 papers, e.g. CVPR 2018 and CVPR 2019 |
| WCT | 256×256 | execution time (sec) | 0.62 – 0.92 | 2 papers, e.g. CVPR 2018 and CVPR 2019 |
| AdaIN | 100 test images | Content loss | 11.43 – 14.42 | 2 papers, e.g. NeurIPS 2019 and AAAI 2020 |
What reviewers will expect
Datasets and baselines named in the 20 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- WikiArt · 5 of 20 papers
- MS-COCO · 3 of 20 papers
Baselines
- StyleID · 8 of 20 papers
- InstantStyle · 7 of 20 papers
- StyTr2 · 6 of 20 papers
- StyleShot · 5 of 20 papers
- DiffuseIT · 4 of 20 papers
- CSGO · 4 of 20 papers
- InST · 3 of 20 papers
- AesPA-Net · 3 of 20 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 57 papers; from OpenAlex and the papers' affiliations).
- Zhejiang University · 7 papers
- Adobe · 4 papers
- Chinese Academy of Sciences · 3 papers
- Jilin University · 3 papers
- Alibaba Group · 2 papers
- Chongqing University · 2 papers
- Kuaishou Technology · 2 papers
- Nanjing University · 2 papers
- Engineering Research Center of Knowledge-Driven Human-Machine Intelligence · 2 papers
- Korea Advanced Institute of Science and Technology · 2 papers
Use it
Download BibTeX of 39 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
- Image-to-image translation (127 papers)
- Texture and sketch synthesis (130 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).