Research map / Image generation / Editing and style transfer
Exploring Stroke-Level Modifications for Scene Text Editing
Yadong Qu, Qingfeng Tan, Hongtao Xie, Jianjun Xu, YuXin Wang, Yongdong Zhang
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 Stroke-Level Modification (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 70.395 -> 76.790 | Improved |
| add recognizer supervision (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 68.375 -> 76.790 | Improved |
| add Background Filtering (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 72.362 -> 76.790 | Improved |
| add Style Augmentation (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 26.408 -> 76.790 | Improved |
| block gradient propagation between BRM and TMM decoders (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 25.670 -> 76.790 | Improved |
| apply SLM in TMM during training (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 76.790 -> 70.032 | Hurt |
| add real images to semi-supervised training (1 image per batch) (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 68.906 -> 75.353 | Improved |
| increase real images per batch from 1 to 2 (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 75.353 -> 76.790 | Improved |
| increase real images per batch from 2 to 4 (ablation) | MOSTEL · Tamper-Scene | SeqAcc: 76.790 -> 68.220 | Hurt |
| train with SLM only in BRM rather than also in TMM (ablation) | MOSTEL · Tamper-Syn2k | FID: 30.66 -> 29.48 | Improved |
| use MOSTEL instead of SwapText | scene text editing · Tamper-Syn2k | FID: 35.62 -> 29.48 | Improved |
| use MOSTEL instead of SwapText | scene text editing · Tamper-Scene | SeqAcc: 60.634 -> 76.790 | 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.
- SwapText — SwapText: Image Based Texts Transfer in Scenes (CVPR 2020) · better on 5 of 5 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
- TextCtrl: Diffusion-based Scene Text Editing with Prior Guidance Control (NeurIPS 2024 · similarity 0.89)
- ViTeX-Bench: Benchmarking High-Fidelity Video Scene Text Editing (NeurIPS 2026 · similarity 0.87)
- DreamText: High Fidelity Scene Text Synthesis (CVPR 2025 · similarity 0.87)
- SteerVTE: Seamless Video Text Editing with Style and Glyph Control (NeurIPS 2026 · similarity 0.86)
- Text-Driven Image Editing via Learnable Regions (CVPR 2024 · similarity 0.86)
- BARET: Balanced Attention Based Real Image Editing Driven by Target-Text Inversion (AAAI 2024 · similarity 0.86)
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.
