Research map / Vision-language and multimodal / Human and embodied modeling
Focusing Attention: Towards Accurate Text Recognition in Natural Images
Zhanzhan Cheng, Fan Bai, Yunlu Xu, Gang Zheng, Shiliang Pu, Shuigeng Zhou
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 the focusing network (FN) to the attention network (ablation) | AN with ResNet encoder · IIIT5k (lexicon-free) | accuracy (%): 83.7 -> 87.4 | Improved |
| add the focusing network (FN) to the attention network (ablation) | AN with ResNet encoder · IC15 (lexicon-free) | accuracy (%): 63.3 -> 70.6 | Improved |
| add the focusing network (FN) to the attention network (ablation) | AN with ResNet encoder · IC13 (lexicon-free) | accuracy (%): 89.4 -> 93.3 | Improved |
| add the focusing network (FN) to the attention network (ablation) | AN with ResNet encoder · IIIT5k (lexicon-free) | total normalized edit distance: 188.9 -> 145.2 | Improved |
| add the focusing network (FN) to the attention network (ablation) | AN with Shi et al. encoder · IIIT5k (lexicon-free) | accuracy (%): 79.5 -> 81.7 | Improved |
| add the focusing network (FN) to the attention network (ablation) | AN with Shi et al. encoder · IC15 (lexicon-free) | accuracy (%): 58.2 -> 61.0 | Improved |
| add the focusing network (FN) to the attention network (ablation) | AN with Shi et al. encoder · IIIT5k (lexicon-free) | total normalized edit distance: 248.0 -> 213.4 | Improved |
| replace the Shi et al. encoder with the 32-layer ResNet-based encoder (ablation) | AN · IIIT5k (lexicon-free) | accuracy (%): 79.5 -> 83.7 | Improved |
| raise lambda from 0.001 to 0.01 (ablation) | FAN · IC13 (lexicon-free) | accuracy (%): 92.7 -> 93.3 | Improved |
| raise lambda from 0.01 to 0.1 (ablation) | FAN · IC13 (lexicon-free) | accuracy (%): 93.3 -> 92.1 | Hurt |
| raise lambda from 0.01 to 0.1 (ablation) | FAN · IC15 (lexicon-free) | accuracy (%): 70.6 -> 70.7 | No clear effect |
| add pixel-level character labels for 1% of training samples (ablation) | FAN · IC15 (lexicon-free) | accuracy (%): 63.3 -> 70.0 | Improved |
Closest papers by meaning
- Scene Text Recognition from Two-Dimensional Perspective (AAAI 2019 · similarity 0.88)
- Single Shot Text Detector With Regional Attention (ICCV 2017 · similarity 0.87)
- An End-to-End TextSpotter With Explicit Alignment and Attention (CVPR 2018 · similarity 0.87)
- Text-Guided Attention Model for Image Captioning (AAAI 2017 · similarity 0.87)
- Decoupled Attention Network for Text Recognition (AAAI 2020 · similarity 0.87)
- ASTER: An Attentional Scene Text Recognizer with Flexible Rectification (TPAMI 2018 · similarity 0.87)
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.
