Research map / Detection and segmentation / CNN features
Verisimilar Image Synthesis for Accurate Detection and Recognition of Texts in Scenes
Fangneng Zhan, Shijian Lu, Chuhui Xue
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 semantic coherence to random embedding (ablation) | synthesis, EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 83.08 -> 84.19 | Improved |
| add saliency guidance to random embedding (ablation) | synthesis, EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 83.08 -> 84.72 | Improved |
| add adaptive text appearance to random embedding (ablation) | synthesis, EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 83.08 -> 83.40 | Improved |
| add adaptive text appearance to SC+SG (ablation) | synthesis, EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 85.94 -> 86.29 | Improved |
| add saliency guidance to SC+ATA (ablation) | synthesis, EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 85.10 -> 86.29 | Improved |
| add semantic coherence to SG+ATA (ablation) | synthesis, EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 85.72 -> 86.29 | Improved |
| add 1K random-embedding synthetic images to real data (ablation) | EAST on ICDAR2013 · ICDAR2013 detection | precision (%): 85.56 -> 84.49 | Hurt |
| increase synthetic images from 1K to 10K (ablation) | EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 86.29 -> 88.25 | Improved |
| add 10K synthetic images to real training data | EAST on ICDAR2015 · ICDAR2015 detection | F-score (%): 79.7 -> 81.9 | Improved |
| add 10K synthetic images to real training data | EAST on MSRA-TD500 · MSRA-TD500 detection | F-score (%): 73.4 -> 78.6 | Improved |
| use our synthesized images instead of Gupta et al. at 10K | EAST on ICDAR2013 · ICDAR2013 detection | F-measure (%): 87.74 -> 88.25 | Improved |
| add 5M synthetic word images to real data | CRNN on ICDAR2013 recognition · ICDAR2013 recognition (no lexicon) | correctly recognized words (%): 31.2 -> 87.1 | Improved |
Closest papers by meaning
- TextNeRF: A Novel Scene-Text Image Synthesis Method based on Neural Radiance Fields (CVPR 2024 · similarity 0.89)
- TextSSR: Diffusion-based Data Synthesis for Scene Text Recognition (ICCV 2025 · similarity 0.89)
- DreamText: High Fidelity Scene Text Synthesis (CVPR 2025 · similarity 0.87)
- Image-Text Co-Decomposition for Text-Supervised Semantic Segmentation (CVPR 2024 · similarity 0.86)
- Kernel Adaptive Convolution for Scene Text Detection via Distance Map Prediction (CVPR 2024 · similarity 0.86)
- Layout-Agnostic Scene Text Image Synthesis with Diffusion Models (CVPR 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.
