Research map / Vision-language and multimodal / Human and embodied modeling
Primitive Representation Learning for Scene Text Recognition
Ruijie Yan, Liangrui Peng, Shanyu Xiao, Gang Yao
Check 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 |
|---|---|---|---|
| Integrate primitive visual text representations into 2D attention | Baseline2D to form PREN2D · IC15, SVTP, and CUTE | word accuracy (%): 81.9 -> 83.0; 86.0 -> 87.6; 89.9 -> 91.7 | Improved |
| Integrate primitive visual text representations into 2D attention | Baseline2D to form PREN2D · Chinese horizontal and vertical | word accuracy (%): 82.2 -> 82.6; 86.8 -> 87.4 | Improved |
| Use parallel primitive representation learning rather than CNN-LSTM-CTC | English scene text recognition · IIIT5k, SVT, IC03, IC13, IC15, SVTP, CUTE | word accuracy (%) | Mixed |
| Combine pooling and weighted feature aggregators (ablation) | PREN · IIIT5k, IC03, IC13, SVTP, CUTE | word accuracy (%): 91.8; 93.9; 94.7; 81.7; 81.3 | Improved |
| Use weighted aggregators instead of pooling aggregators (ablation) | PREN · IIIT5k, IC03, IC13 | word accuracy (%): 91.1 -> 90.0; 93.5 -> 92.0; 94.7 -> 93.2 | Hurt |
| Use weighted aggregators instead of pooling aggregators (ablation) | PREN · SVTP and CUTE | word accuracy (%): 79.2 -> 79.5; 77.4 -> 77.4 | Mixed |
| Vary the number of primitive representations; five performed best on the listed datasets (ablation) | PREN · IIIT5k, IC13, SVTP, CUTE | word accuracy (%) | Mixed |
Closest papers by meaning
- IPAD: Iterative, Parallel, and Diffusion-Based Network for Scene Text Recognition (IJCV 2025 · similarity 0.88)
- Vision-Language Pre-Training for Boosting Scene Text Detectors (CVPR 2022 · similarity 0.87)
- Scene Text Recognition from Two-Dimensional Perspective (AAAI 2019 · similarity 0.87)
- Language Matters: A Weakly Supervised Vision-Language Pre-training Approach for Scene Text Detection and Spotting (ECCV 2022 · similarity 0.87)
- Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition (AAAI 2019 · similarity 0.87)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.87)
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.
