Research map / Image generation / Editing and style transfer
Texture Networks: Feed-forward Synthesis of Textures and Stylized Images
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 iterative optimization with a feed-forward generator network for texture synthesis | texture synthesis · texture synthesis vs Gatys et al. | generation speed: 500x faster | Helped |
| feed-forward generation instead of backprop-based optimization | texture synthesis · 256x256 samples | memory use: 170 MB vs 1100 MB | Helped |
| multi-scale generator architecture instead of simple chain of conv/upsampling layers (ablation) | generator network · texture synthesis | texture loss and perceptual quality | Helped |
| batch normalization inserted before concatenation layers (ablation) | generator network · texture training | training quality | Helped |
| strided full-convolution upsampling instead of nearest-neighbour upsampling (ablation) | generator network · texture synthesis | visual quality | Hurt |
| train style transfer on a small pool of about 16 content images instead of many (ablation) | style transfer generator · stylization | qualitative result quality | Helped |
| scale the input noise magnitude at test time | style transfer generator · stylization | style/content balance | Helped |
| feed-forward stylization instead of optimization-based style transfer | style transfer · stylization vs Gatys et al. | perceptual quality | Mixed |
Closest papers by meaning
- Diversified Texture Synthesis With Feed-Forward Networks (CVPR 2017 · similarity 0.94)
- TextureGAN: Controlling Deep Image Synthesis With Texture Patches (CVPR 2018 · similarity 0.91)
- GramGAN: Deep 3D Texture Synthesis From 2D Exemplars (NeurIPS 2020 · similarity 0.91)
- Fast Texture Synthesis via Pseudo Optimizer (CVPR 2020 · similarity 0.90)
- Improved Texture Networks: Maximizing Quality and Diversity in Feed-Forward Stylization and Texture Synthesis (CVPR 2017 · similarity 0.90)
- Texture Mixer: A Network for Controllable Synthesis and Interpolation of Texture (CVPR 2019 · 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.