Research map / Image generation / Editing and style transfer
Neural Preset for Color Style Transfer
Zhanghan Ke, Yuhao Liu, Lei Zhu, Nanxuan Zhao, Rynson W.H. Lau
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 Neural Preset instead of the strongest-ranked baseline, WCT² | color style transfer · user study | average ranking: 2.67 -> 1.81 | Improved |
| use Neural Preset instead of PhotoWCT² | color style transfer · 2K (2560 × 1440), RTX3090 | GPU inference time: 0.447\,s -> 0.016\,s | Improved |
| use Neural Preset instead of PhotoWCT² | color style transfer · 8K (7680 × 4320), RTX3090 | GPU memory: OOM -> 1.96\,GB | Improved |
| use DNCM instead of the strongest baseline, CDTNet | image harmonization · iHarmony4 | MSE: 23.75 -> 24.31 | Hurt |
| use DNCM instead of CDTNet | image harmonization · iHarmony4 | PSNR: 38.23 -> 37.97 | Hurt |
| use DNCM instead of Learnable 3D LUT | image color enhancement · MIT-Adobe FiveK | SSIM: 0.8636 -> 0.8697 | Improved |
| use DNCM instead of Learnable 3D LUT | image color enhancement · MIT-Adobe FiveK | PSNR: 23.17 -> 23.12 | Hurt |
| replace DNCM with two CNNs for color mapping (ablation) | Neural Preset self-supervised pipeline · color style transfer | training convergence | Hurt |
| add the consistency loss L_con (ablation) | Neural Preset · color style transfer | training convergence | Improved |
| reduce DNCM parameter dimension from k=16 to k=2 (ablation) | Neural Preset · color style transfer validation set | style similarity: 0.746 -> 0.128 | Hurt |
| increase DNCM parameter dimension from k=16 to k=32 (ablation) | Neural Preset · color style transfer validation set | style similarity: 0.746 -> 0.769 | Improved |
| increase DNCM patch size from 512 to 1024 (ablation) | Neural Preset · 8K color style transfer | GPU memory: 1.96\,GB -> 2.14\,GB | Hurt |
Closest papers by meaning
- Universal Style Transfer via Feature Transforms (NeurIPS 2017 · similarity 0.89)
- Controlling Perceptual Factors in Neural Style Transfer (CVPR 2017 · similarity 0.88)
- Arbitrary Style Transfer in Real-Time With Adaptive Instance Normalization (ICCV 2017 · similarity 0.87)
- Image-To-Image Translation via Group-Wise Deep Whitening-And-Coloring Transformation (CVPR 2019 · similarity 0.87)
- Understanding Generalized Whitening and Coloring Transform for Universal Style Transfer (ICCV 2019 · similarity 0.87)
- Style Injection in Diffusion: A Training-free Approach for Adapting Large-scale Diffusion Models for Style Transfer (CVPR 2024 · similarity 0.85)
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.
