Research map / Image generation / Editing and style transfer
Towards compact reversible image representations for neural style transfer
Xiyao Liu, Siyu Yang, Jian Zhang, Gerald Schaefer, Jiya Li, Xunli FAN, Songtao Wu, Hui Fang
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 InfoMinRev module, Barlow twins loss and JSD style loss together (ablation) | flow baseline · style transfer, 1200 images | SSIM: 0.372 -> 0.496 | Improved |
| add InfoMinRev module (ablation) | full model without IMR · style transfer, 1200 images | SSIM: 0.477 -> 0.496 | Improved |
| add Barlow twins channel-wise loss (ablation) | full model without LB · style transfer, 1200 images | SSIM: 0.459 -> 0.496 | Improved |
| use Jensen-Shannon divergence style loss instead of mean-variance L2 (ablation) | full model · style transfer, 1200 images | style loss: 0.836 -> 0.652 | Improved |
| add mutual information minimisation loss L_info (ablation) | InfoMinRev flow · intermediate feature information distribution | information in feature maps (bits) | Improved |
| remove IMR and Barlow loss together (ablation) | full model · style transfer, 1200 images | style loss: 0.652 -> 0.566 | Mixed |
| propose method versus SANet and CAP-VSTNet on style loss | Ours · style transfer, 1200 images | style loss: 0.587 (SANet), 0.608 (CAP-VSTNet) vs 0.652 | Hurt |
| propose method versus CAP-VSTNet and ArtFlow on runtime | Ours · 512x512 images | time (s): 0.096 -> 0.081 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ArtFlow — ArtFlow: Unbiased Image Style Transfer via Reversible Neural Flows (CVPR 2021) · better on 2 of 2 shared results
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
- ArtFlow: Unbiased Image Style Transfer via Reversible Neural Flows (CVPR 2021 · similarity 0.91)
- Style-Aware Normalized Loss for Improving Arbitrary Style Transfer (CVPR 2021 · similarity 0.90)
- Rethinking and Improving the Robustness of Image Style Transfer (CVPR 2021 · similarity 0.90)
- StyleBank: An Explicit Representation for Neural Image Style Transfer (CVPR 2017 · similarity 0.89)
- Separating Style and Content for Generalized Style Transfer (CVPR 2018 · similarity 0.89)
- A Style-Aware Content Loss for Real-time HD Style Transfer (ECCV 2018 · similarity 0.89)
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.
