Research map / Vision-language and multimodal / Human and embodied modeling
Vision-Language Pre-Training for Boosting Scene Text Detectors
Sibo Song, Jianqiang Wan, Zhibo Yang, Jun Tang, Wenqing Cheng, Xiang Bai, Cong 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 |
|---|---|---|---|
| Use VLPT-STD pre-training instead of STKM | PSENet · Total-Text | F-measure: 84.3 -> 86.1 | Improved |
| Use VLPT-STD pre-training instead of STKM | EAST · MSRA-TD500 | F-measure: 80.0 -> 82.2 | Improved |
| Use VLPT-STD pre-training instead of STKM | DB · MSRA-TD500 | F-measure: 85.9 -> 88.5 | Improved |
| Use the proposed FPN instead of the original FPN (ablation) | VLPT-STD fine-tuned with PSENet · CTW1500 | F-measure: 82.7 -> 83.3 | Improved |
| Use the proposed FPN instead of the original FPN (ablation) | VLPT-STD fine-tuned with EAST · ICDAR2015 | F-measure: 88.0 -> 88.3 | Improved |
| Add multi-head cross-attention rather than using only self-attention (ablation) | VLPT-STD · CTW1500 | F-measure: 83.1 -> 83.3 | Improved |
| Add MLM alone rather than using SynthText detector pre-training (ablation) | PSENet · ICDAR2015 | F-measure: 81.3 -> 84.5 | Improved |
| Add WIP alone rather than using SynthText detector pre-training (ablation) | PSENet · ICDAR2015 | F-measure: 81.3 -> 83.1 | Improved |
| Add ITC alone rather than using SynthText detector pre-training (ablation) | EAST · MSRA-TD500 | F-measure: 82.0 -> 79.3 | Hurt |
| Add WIP to ITC + MLM (ablation) | PSENet · Total-Text | F-measure: 85.6 -> 86.1 | Improved |
| Add WIP to ITC + MLM (ablation) | EAST · ICDAR2015 | F-measure: 87.5 -> 88.3 | Improved |
| Add TextOCR to SynthText pre-training (ablation) | PSENet with text supervision · ICDAR2015 | F-measure: 84.33 -> 85.07 | Improved |
Closest papers by meaning
- Language Matters: A Weakly Supervised Vision-Language Pre-training Approach for Scene Text Detection and Spotting (ECCV 2022 · similarity 0.92)
- Image as a Language: Revisiting Scene Text Recognition via Balanced, Unified and Synchronized Vision-Language Reasoning Network (AAAI 2024 · similarity 0.91)
- UNITER: UNiversal Image-TExt Representation Learning (ECCV 2020 · similarity 0.90)
- PreSTU: Pre-Training for Scene-Text Understanding (ICCV 2023 · similarity 0.90)
- Revisiting the Role of Language Priors in Vision-Language Models (ICML 2024 · similarity 0.89)
- Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models (NeurIPS 2025 · 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.
