Research map / Vision-language and multimodal / Human and embodied modeling
ABINet++: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Spotting
Shancheng Fang, Zhendong Mao, Hongtao Xie, Yuxin Wang, Chenggang Clarence Yan, 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 |
|---|---|---|---|
| use position attention and 3 Transformer layers (LV) over small vision model (ablation) | vision model · STR average | avg accuracy: 88.8 -> 89.8 | Improved |
| pre-train vision model (PVM) (ablation) | LV+BCN · STR average | avg accuracy: 91.7 -> 92.3 | Improved |
| add LM pre-training on WikiText-103 (PLM_out) (ablation) | LV+BCN · STR average | avg accuracy: 92.3 -> 92.5 | Improved |
| allow gradient flow between vision and language models (ablation) | LV+BCN · STR average | avg accuracy: 92.3 -> 91.4 | Hurt |
| bidirectional BCN instead of unidirectional SRN-U (ablation) | SV · STR average | avg accuracy: 90.2 -> 91.0 | Improved |
| BCN instead of SRN (ablation) | LV · STR average | avg accuracy: 91.9 -> 92.3 | Improved |
| BCN instead of SRN in spelling correction (ablation) | language model · text spelling correction | word accuracy: 27.6 -> 41.9 | Improved |
| BCN instead of BERT (ablation) | LV · STR average | avg accuracy: 92.3 -> 92.3 | No clear effect |
| iterate BCN 3 times (ablation) | SV+BCN · STR average | avg accuracy: 91.0 -> 91.4 | Improved |
| iterate BCN 3 times (ablation) | LV+BCN · STR average | avg accuracy: 92.3 -> 92.6 | Improved |
| add language model fine-tuned on target data (ablation) | spotting · Total-Text | Hmean: 70.2 -> 76.3 | Improved |
| add language model fine-tuned on target data (ablation) | spotting · SCUT-CTW1500 | Hmean: 53.5 -> 59.4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- RobustScanner — RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition (ECCV 2020) · better on 0 of 1 shared result
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
- Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition (CVPR 2021 · similarity 0.92)
- SPTS v2: Single-Point Scene Text Spotting (TPAMI 2023 · similarity 0.89)
- SwinTextSpotter: Scene Text Spotting via Better Synergy Between Text Detection and Text Recognition (CVPR 2022 · similarity 0.89)
- Language Matters: A Weakly Supervised Vision-Language Pre-training Approach for Scene Text Detection and Spotting (ECCV 2022 · similarity 0.89)
- Image as a Language: Revisiting Scene Text Recognition via Balanced, Unified and Synchronized Vision-Language Reasoning Network (AAAI 2024 · similarity 0.88)
- 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.
