Research map / Image generation / Editing and style transfer
Photorealistic Style Transfer via Wavelet Transforms
Jaejun Yoo, Youngjung Uh, Sanghyuk Chun, Byeongkyu Kang, Jung-Woo Ha
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 |
|---|---|---|---|
| replace max-pooling with Haar wavelet pooling and unpooling (no post-processing) | WCT2 vs PhotoWCT · photorealistic stylization | user preference (most preferred): 62.21% | Improved |
| use single-pass progressive stylization instead of multi-level pipeline with removal of post-processing | WCT2 · 1024x1024 runtime vs DPST | runtime (s): 3887.8 -> 4.7 | Improved |
| use wavelet pooling instead of multi-level max-pool PhotoWCT: memory | WCT2 · GPU memory | GPU memory: 51% of PhotoWCT GPU memory | Improved |
| add wavelet pooling instead of split pooling (ablation) | WCT2 · photorealistic stylization (qualitative) | fine detail preservation | Improved |
| add wavelet pooling instead of learnable strided conv pooling (ablation) | WCT2 · photorealistic stylization (qualitative) | content preservation and style transfer | Improved |
| use concatenation instead of summation in wavelet unpooling (ablation) | WCT2 · photorealistic stylization (qualitative) | spatial detail and parameters: 1.80x parameters of sum-version | Mixed |
| add multi-level stylization on top of WCT2 (ablation) | WCT2 · statistics SSIM vs style loss | style strength vs SSIM | Mixed |
| add progressive WCT at decoder layers conv3_2, conv2_2, conv1_2 (ablation) | WCT2 · stylization strength | stylization strength | Mixed |
| stylize only LL component instead of all wavelet components (ablation) | WCT2 · stylization of edges | edge stylization | Mixed |
| add post-processing smoothing to PhotoWCT | PhotoWCT · SSIM vs style loss | visual quality | Improved |
Closest papers by meaning
- Ultrafast Photorealistic Style Transfer via Neural Architecture Search (AAAI 2020 · similarity 0.90)
- Joint Bilateral Learning for Real-time Universal Photorealistic Style Transfer (ECCV 2020 · similarity 0.90)
- A Content Transformation Block for Image Style Transfer (CVPR 2019 · similarity 0.90)
- Deep Photo Style Transfer (CVPR 2017 · similarity 0.89)
- A Closed-form Solution to Photorealistic Image Stylization (ECCV 2018 · similarity 0.89)
- Non-Local Representation Based Mutual Affine-Transfer Network for Photorealistic Stylization (TPAMI 2021 · 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.
