Research map / Vision-language and multimodal / Human and embodied modeling
From Two to One: A New Scene Text Recognizer With Visual Language Modeling Network
Yuxin Wang, Hongtao Xie, Shancheng Fang, Jing Wang, Shenggao Zhu, 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 MLM occluded training (ratio 1:1) (ablation) | VRM-2L baseline · IC15 | accuracy: 79.8 -> 81.8 | Improved |
| add MLM occluded training (ratio 1:1) (ablation) | VRM-2L baseline · SVTP | accuracy: 81.1 -> 83.7 | Improved |
| add MLM occluded training (ratio 1:1) (ablation) | VRM-2L baseline · CUTE80 | accuracy: 85.8 -> 88.2 | Improved |
| add MLM occluded training (ratio 1:1) (ablation) | VRM-2L baseline · IIIT5K | accuracy: 94.5 -> 95.4 | Improved |
| raise occluded ratio from 1:1 to 2:1 (ablation) | MLM · SVT | accuracy: 91.0 -> 90.0 | Hurt |
| use both WCL branches instead of Mas only (ablation) | MLM · SVT | accuracy: 89.8 -> 91.0 | Improved |
| use both WCL branches instead of Rem only (ablation) | MLM · SVTP | accuracy: 82.2 -> 83.7 | Improved |
| replace dropout/cutout with MLM masking (ablation) | masking method · 6-benchmark average | accuracy: 89.0 -> 90.2 | Improved |
| random cutout instead of no masking (ablation) | baseline · 6-benchmark average | accuracy: 88.8 -> 89.0 | No clear effect |
| use three transformer units instead of two in VRM (ablation) | VRM · IC15 | accuracy: 81.8 -> 83.7 | Improved |
| use three transformer units instead of two in VRM (ablation) | VRM · CUTE80 | accuracy: 88.2 -> 88.5 | Improved |
| replace separate language model with VisionLAN (ablation) | baseline + Transformer LM · OST | average accuracy: 58.2 -> 60.3 | Improved |
Closest papers by meaning
- Image as a Language: Revisiting Scene Text Recognition via Balanced, Unified and Synchronized Vision-Language Reasoning Network (AAAI 2024 · similarity 0.91)
- Language Matters: A Weakly Supervised Vision-Language Pre-training Approach for Scene Text Detection and Spotting (ECCV 2022 · similarity 0.90)
- Linguistics-aware Masked Image Modeling for Self-supervised Scene Text Recognition (CVPR 2025 · similarity 0.90)
- Visual Semantics Allow for Textual Reasoning Better in Scene Text Recognition (AAAI 2022 · similarity 0.90)
- Multi-modal Text Recognition Networks: Interactive Enhancements between Visual and Semantic Features (ECCV 2022 · similarity 0.89)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.88)
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.
