Research map / Image generation / Super-resolution
A Text Attention Network for Spatial Deformation Robust Scene Text 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 global TP Interpreter cross-attention instead of local TP-to-image fusion (ablation) | TATT · TextZoom | CRNN accuracy: 41.4% -> 52.6% | Helped |
| replace TP Interpreter with Pixel-Shuffle alignment and addition (ablation) | TATT · TextZoom | CRNN accuracy: 52.6% -> 46.2% | Hurt |
| add text structure consistency loss (ablation) | TATT · TextZoom | CRNN accuracy: 51.6% -> 52.6% | Helped |
| add text structure consistency loss (ablation) | TSRN · TextZoom | CRNN accuracy: 41.4% -> 45.7% | Helped |
| add text structure consistency loss (ablation) | TBSRN · TextZoom | CRNN accuracy: 48.1% -> 49.4% | Helped |
| add text structure consistency loss (ablation) | TPGSR · TextZoom | CRNN accuracy: 49.8% -> 50.6% | Helped |
| use recurrent positional encoding instead of fixed positional encoding in the decoder (ablation) | TATT · TextZoom SR images | CRNN accuracy: 50.5% -> 52.6% | Helped |
| use learnable positional encoding instead of recurrent positional encoding in the decoder (ablation) | TATT · TextZoom SR images | CRNN accuracy: 52.6% -> 50.8% | Hurt |
| use global text-semantic guidance and TSC-trained reconstruction | TATT · TextZoom test (avg) | PSNR: PCAN 21.49; TATT 21.52 | Helped |
| use global text-semantic guidance and TSC-trained reconstruction | TATT · TextZoom test (avg) | MORAN accuracy: TPGSR-3 60.5%; TATT 59.5% | Mixed |
| use TATT for spatially-deformed text reconstruction | TATT · TextZoom spatially-deformed samples | CRNN accuracy: 40.2% -> 43.8% | Helped |
Closest papers by meaning
- Gradient-Based Graph Attention for Scene Text Image Super-resolution (AAAI 2023 · similarity 0.90)
- Scene Text Recognition from Two-Dimensional Perspective (AAAI 2019 · similarity 0.89)
- Scene Text Telescope: Text-Focused Scene Image Super-Resolution (CVPR 2021 · similarity 0.89)
- Text-Aware Real-World Image Super-Resolution via Diffusion Model with Joint Segmentation Decoders (NeurIPS 2025 · similarity 0.86)
- TextNeRF: A Novel Scene-Text Image Synthesis Method based on Neural Radiance Fields (CVPR 2024 · similarity 0.85)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · 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.