Research map / Image generation / Super-resolution
Scene Text Telescope: Text-Focused Scene Image Super-Resolution
Check 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 |
|---|---|---|---|
| Use STT instead of TSRN for super-resolution | CRNN on TextZoom · TextZoom test | recognition accuracy, average: 41.4% -> 48.1% | Helped |
| Use STT instead of TSRN for super-resolution | ASTER on TextZoom · TextZoom test | recognition accuracy, average: 58.3% -> 60.1% | Helped |
| Use STT instead of TSRN for super-resolution | MORAN on TextZoom · TextZoom test | recognition accuracy, average: 55.4% -> 58.4% | Helped |
| Add 2-D positional encoding to TBSRN (ablation) | TBSRN-5 on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 37.9% -> 43.2% | Helped |
| Use five TBSRN blocks rather than fewer or more tested block counts (ablation) | TBSRN on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 39.8%, 41.4%, 42.1%, 41.7%, 43.2%, 42.3% | Mixed |
| Add Position-Aware Module with λPOS 10 (ablation) | TBSRN-5 on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 43.2% -> 45.6% | Helped |
| Vary λPOS from 0 across tested settings (ablation) | TBSRN-5 on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 43.2%, 42.3%, 44.0%, 45.6%, 45.0% | Mixed |
| Add Content-Aware Module (ablation) | TBSRN with Position-Aware Module on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 45.6% -> 48.1% | Helped |
| Use weighted cross-entropy for confusable characters instead of vanilla cross-entropy (ablation) | Content-Aware Module on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 47.1% -> 48.1% | Helped |
| Vary λCON across tested values (ablation) | Content-Aware Module on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 45.6%, 46.6%, 47.5%, 47.1%, 48.1%, 46.2%, 47.2%, 46.1%, 46.3% | Mixed |
| Add pixel-wise supervision LPSM (ablation) | STT on TextZoom · TextZoom test | recognition accuracy (CRNN), average: 36.6% -> 48.1% | Helped |
| Preprocess Gaussian-blurred benchmark images with STT | CRNN on IC13, IC15, and CT80 · IC13, IC15, and CT80; Gaussian blur radius 3 | recognition accuracy: 43.0% -> 50.8%; 5.2% -> 8.4%; 26.4% -> 33.7% | Helped |
Closest papers by meaning
- Text-Aware Real-World Image Super-Resolution via Diffusion Model with Joint Segmentation Decoders (NeurIPS 2025 · similarity 0.87)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.87)
- Boosting Semi-Supervised Scene Text Recognition via Viewing and Summarizing (NeurIPS 2024 · similarity 0.85)
- Beyond Image Super-Resolution for Image Recognition with Task-Driven Perceptual Loss (CVPR 2024 · similarity 0.84)
- CLIP is Almost All You Need: Towards Parameter-Efficient Scene Text Retrieval without OCR (CVPR 2025 · similarity 0.84)
- Accurate Scene Text Recognition with Efficient Model Scaling and Cloze Self-Distillation (CVPR 2025 · similarity 0.84)
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.