Research map / Image generation / Editing and style transfer
TET-GAN: Text Effects Transfer via Stylization and Destylization
Shuai Yang, Jiaying Liu, Wenjing Wang, Zongming Guo
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 reconstruction and feature losses (ablation) | TET-GAN · Text effects transfer | visual transfer quality | Improved |
| Add feature loss to guide content-feature extraction (ablation) | TET-GAN · Text effects transfer | glyph disentanglement and artifact quality | Improved |
| Add distribution-aware text-image preprocessing and text-effects augmentation (ablation) | TET-GAN · Text effects transfer | texture placement quality | Improved |
| Replace Batch Normalization with Instance Normalization in the text-effects content encoder | TET-GAN destylization subnetwork · Destylization | destylization quality | Improved |
| Pretrain on the text-effects dataset before one-shot finetuning | TET-GAN · One-shot supervised style transfer | stylization quality | Improved |
| Use the proposed self-stylization finetuning scheme | TET-GAN · One-shot supervised style transfer | inference time: one minute per image -> 20ms per image | Improved |
| Add common style autoencoder reconstruction or cycle-consistency losses (ablation) | TET-GAN stylization subnetwork · Text effects transfer | stylization results | No clear effect |
| Use pixel-domain patch matching | T-Effect · Text effects transfer | color continuity | Hurt |
| Use greedy deep-feature patch matching | Neural Doodles · Text effects transfer | glyph shape preservation | Hurt |
| Train TET-GAN from scratch rather than pretraining it on the proposed dataset (ablation) | TET-GAN · One-shot supervised style transfer | stylization quality | Hurt |
Closest papers by meaning
- TE141K: Artistic Text Benchmark for Text Effect Transfer (TPAMI 2020 · similarity 0.93)
- FET-GAN: Font and Effect Transfer via K-shot Adaptive Instance Normalization (AAAI 2020 · similarity 0.90)
- Multi-Content GAN for Few-Shot Font Style Transfer (CVPR 2018 · similarity 0.90)
- TextCtrl: Diffusion-based Scene Text Editing with Prior Guidance Control (NeurIPS 2024 · similarity 0.86)
- CSGO: Content-Style Composition in Text-to-Image Generation (NeurIPS 2025 · similarity 0.85)
- SigStyle: Signature Style Transfer via Personalized Text-to-Image Models (AAAI 2025 · 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.
