Research map / Detection and segmentation / CNN features
Optimal Boxes: Boosting End-to-End Scene Text Recognition by Adjusting Annotated Bounding Boxes via Reinforcement Learning
Jingqun Tang, Wenming Qian, Luchuan Song, Xiena Dong, Lan Li, Xiang Bai
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 |
|---|---|---|---|
| train with BoxDQN-adjusted boxes | EAST+CRNN · IC15 | F-score: 82.2 -> 83.9 | Improved |
| train with BoxDQN-adjusted boxes | EAST+CRNN · MLT19 | F-score: 53.9 -> 56.9 | Improved |
| train with BoxDQN-adjusted boxes | two-step pipelines · IC13, IC15, MLT19, ReCTS | F-score gain: at least 1.6% | Improved |
| train with BoxDQN-adjusted boxes | ABCNet · MLT19 | F-score: 56.2 -> 59.5 | Improved |
| train with BoxDQN-adjusted boxes | MTS-V3 · MLT19 | F-score: 61.2 -> 64.0 | Improved |
| train with BoxDQN-adjusted boxes | FOTS · IC15 | F-score: 81.5 -> 83.3 | Improved |
| adjust pseudo-labels with BoxDQN for synthetic-to-real adaptation | EAST+CRNN · MLT19 (cross-domain) | F-score: 41.3 -> 48.3 | Improved |
| adjust pseudo-labels with BoxDQN for synthetic-to-real adaptation | MTS-V3 · IC15 (cross-domain) | F-score: 71.3 -> 76.0 | Improved |
| use BoxDQN instead of grid search (ablation) | EAST+CRNN · IC15 | F-score: 82.8 -> 83.9 | Improved |
| use BoxDQN instead of grid search (ablation) | MTS-V3 · IC15 | F-score: 83.4 -> 84.5 | Improved |
| add background image and box coordinates to the DQN input via FFM (ablation) | DBNET+RARE · IC15 | F-score: 85.7 -> 86.3 | Improved |
| add background image and box coordinates to the DQN input via FFM (ablation) | MTS-V3 · IC15 | F-score: 83.8 -> 84.5 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ABCNet — ABCNet: Real-Time Scene Text Spotting With Adaptive Bezier-Curve Network (CVPR 2020) · better on 5 of 5 shared results
- FOTS — FOTS: Fast Oriented Text Spotting With a Unified Network (CVPR 2018) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- An End-to-End TextSpotter With Explicit Alignment and Attention (CVPR 2018 · similarity 0.86)
- TextBoxes: A Fast Text Detector with a Single Deep Neural Network (AAAI 2017 · similarity 0.85)
- TAP: Text-Aware Pre-Training for Text-VQA and Text-Caption (CVPR 2021 · similarity 0.85)
- TextOCR: Towards Large-Scale End-to-End Reasoning for Arbitrary-Shaped Scene Text (CVPR 2021 · similarity 0.85)
- ODM: A Text-Image Further Alignment Pre-training Approach for Scene Text Detection and Spotting (CVPR 2024 · similarity 0.85)
- Boosting Semi-Supervised Scene Text Recognition via Viewing and Summarizing (NeurIPS 2024 · similarity 0.84)
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.
