Research map / Detection and segmentation / Weakly supervised detection and segmentation
WeText: Scene Text Detection Under Weak Supervision
Shangxuan Tian, Shijian Lu, Chongshou Li
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 semi-supervised samples mined from FORU (ablation) | light character detector · ICDAR 2013 character detection | F-score: 71.26 -> 73.05 | Improved |
| add weak word-level labels on FORU on top of semi-supervised (ablation) | light character detector · ICDAR 2013 character detection | F-score: 73.05 -> 75.37 | Improved |
| add semi-supervised samples mined from COCO-Text (ablation) | light character detector · ICDAR 2013 character detection | F-score: 71.26 -> 74.91 | Improved |
| add weak labels on COCO-Text on top of semi-supervised (ablation) | light character detector · ICDAR 2013 character detection | F-score: 74.91 -> 78.38 | Improved |
| use larger COCO-Text instead of FORU for weak supervision (ablation) | WeText · ICDAR 2013 character detection | F-score: 75.37 -> 78.38 | Improved |
| add weak supervision from COCO-Text, with text line extraction (ablation) | WeText · ICDAR 2013 text line detection | F-score: 82.3 -> 86.9 | Improved |
| add semi-supervised COCO-Text samples with text line extraction (ablation) | WeText · ICDAR 2013 text line detection | F-score: 82.3 -> 84.2 | Improved |
| add weak supervision from FORU with text line extraction (ablation) | WeText · ICDAR 2013 text line detection | F-score: 82.3 -> 85.4 | Improved |
| replace weak labels with full ground truth on FORU (ablation) | WeText · ICDAR 2013 text line detection | F-score: 85.4 -> 86.3 | Improved |
| add weak supervision from COCO-Text (ablation) | WeText · SWT text line detection | F-score: 53.9 -> 59.8 | Improved |
| add weak supervision from FORU (ablation) | WeText · SWT text line detection | F-score: 53.9 -> 57.1 | Improved |
| second round of iterative sample mining for weakly supervised FORU model (ablation) | WeText · ICDAR 2013 text line detection | F-score: 85.4 -> 86.2 | Improved |
Closest papers by meaning
- Language Matters: A Weakly Supervised Vision-Language Pre-training Approach for Scene Text Detection and Spotting (ECCV 2022 · similarity 0.90)
- WordSup: Exploiting Word Annotations for Character Based Text Detection (ICCV 2017 · similarity 0.89)
- EAST: An Efficient and Accurate Scene Text Detector (CVPR 2017 · similarity 0.88)
- ABINet++: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Spotting (TPAMI 2022 · similarity 0.88)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.87)
- Boosting Semi-Supervised Scene Text Recognition via Viewing and Summarizing (NeurIPS 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.
