Research map / Image generation / Editing and style transfer
AdaAttN: Revisit Attention Mechanism in Arbitrary Neural Style Transfer
Songhua Liu, Tianwei Lin, Dongliang He, Fu Li, Meiling Wang, Xin Li, Zhengxing Sun, Qian Li, Errui Ding
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 |
|---|---|---|---|
| use cosine-similarity attention instead of softmax (ablation) | AdaAttN · Video style transfer (mean of 20 styles) | optical flow error (x1e-2): 5.52 -> 4.09 | Improved |
| add cross-image similarity loss L_is (ablation) | AdaAttN · Video style transfer (mean of 20 styles) | optical flow error (x1e-2): 5.52 -> 5.51 | No clear effect |
| add cross-image similarity loss L_is on top of cosine attention (ablation) | AdaAttN + Cos · Video style transfer (mean of 20 styles) | optical flow error (x1e-2): 4.09 -> 3.91 | Improved |
| replace local feature loss with L2 content loss (ablation) | AdaAttN · Image style transfer | visual quality | Hurt |
| add local feature loss instead of content loss (ablation) | AdaAttN · Image style transfer | visual quality | Improved |
| add global style loss to local feature loss (ablation) | AdaAttN · Image style transfer | color saturation | Improved |
| add shallow features to Q and K (ablation) | AdaAttN · Image style transfer | visual quality | Improved |
| use AdaAttN instead of SANet for video | AdaAttN vs SANet · Video style transfer (mean of 20 styles) | optical flow error (x1e-2): 7.76 -> 5.52 | Improved |
| add cosine attention and L_is to beat Linear and MCCNet | AdaAttN + Cos + L_is · Video style transfer (mean of 20 styles) | optical flow error (x1e-2): 4.42 -> 3.91 | Improved |
| use AdaAttN instead of SANet for speed | AdaAttN vs SANet · Inference time, 512x512 | seconds per image: 0.081 -> 0.112 | Hurt |
Later papers that compared with this method
11 later accepted papers report AdaAttN as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.
- SigStyle: Signature Style Transfer via Personalized Text-to-Image Models (AAAI 2025 · better on 3 of 4 shared results)
- HSI: A Holistic Style Injector for Arbitrary Style Transfer (CVPR 2025 · better on 4 of 4 shared results)
- AesFA: An Aesthetic Feature-Aware Arbitrary Neural Style Transfer (AAAI 2024 · better on 0 of 2 shared results)
- All-to-Key Attention for Arbitrary Style Transfer (ICCV 2023 · better on 5 of 5 shared results)
- QuantArt: Quantizing Image Style Transfer Towards High Visual Fidelity (CVPR 2023 · better on 2 of 2 shared results)
- AesPA-Net: Aesthetic Pattern-Aware Style Transfer Networks (ICCV 2023 · better on 1 of 2 shared results)
- Neural Collage Transfer: Artistic Reconstruction via Material Manipulation (ICCV 2023 · better on 2 of 2 shared results)
- Painterly Image Harmonization in Dual Domains (AAAI 2023 · better on 1 of 1 shared result)
3 more not listed.
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
- Adaptive Convolutions for Structure-Aware Style Transfer (CVPR 2021 · similarity 0.92)
- Arbitrary Style Transfer in Real-Time With Adaptive Instance Normalization (ICCV 2017 · similarity 0.91)
- Arbitrary Style Transfer With Style-Attentional Networks (CVPR 2019 · similarity 0.90)
- Attention-Aware Multi-Stroke Style Transfer (CVPR 2019 · similarity 0.90)
- Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 2024 · similarity 0.88)
- SCSA: A Plug-and-Play Semantic Continuous-Sparse Attention for Arbitrary Semantic Style Transfer (CVPR 2025 · similarity 0.87)
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.
