Research map / Detection and segmentation / CNN features
An End-to-End TextSpotter With Explicit Alignment and Attention
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace RoI pooling with text alignment (ablation) | TextSpotter recognition · ICDAR2015, no lexicon, GT detections | Accuracy (%): 60.7 to 67.6 | Helped |
| replace RoI pooling with RoI alignment (ablation) | TextSpotter recognition · ICDAR2015, no lexicon, GT detections | Accuracy (%): 60.7 to 61.9 | Helped |
| add character supervision (ablation) | Text alignment recognition model · ICDAR2015, no lexicon, GT detections | Accuracy (%): 67.6 to 68.8 | Helped |
| add position embedding (ablation) | Text alignment recognition model · ICDAR2015, no lexicon, GT detections | Accuracy (%): 67.6 to 68.2 | Helped |
| add character supervision and position embedding (ablation) | Text alignment recognition model · ICDAR2015, no lexicon, GT detections | Accuracy (%): 67.6 to 69.5 | Helped |
| jointly train detection and recognition instead of detection without recognition (ablation) | TextSpotter · ICDAR2013 detection, ICDAR standard | F-measure: 0.88 to 0.90 | Helped |
| jointly train detection and recognition instead of detection without recognition (ablation) | TextSpotter · ICDAR2015 detection | F-measure: 0.83 to 0.87 | Helped |
| use explicit character attention supervision (ablation) | Attention-based word recognition · VGG synthetic test set | Character-level accuracy: 0.93 to 0.95 | Helped |
| use explicit character attention supervision (ablation) | Attention-based word recognition · VGG synthetic test set | Word-level accuracy: 0.85 to 0.88 | Helped |
| use the proposed end-to-end spotter instead of Deep TextSpotter | TextSpotter · ICDAR2015 end-to-end, strong lexicon | F-measure: 0.54 -> 0.82 | Helped |
| use the proposed end-to-end spotter instead of Deep TextSpotter | TextSpotter · ICDAR2015 end-to-end, weak lexicon | F-measure: 0.51 -> 0.77 | Helped |
| use the proposed end-to-end spotter instead of Deep TextSpotter | TextSpotter · ICDAR2015 end-to-end, generic lexicon | F-measure: 0.47 -> 0.63 | Helped |
Closest papers by meaning
- Implicit Feature Alignment: Learn To Convert Text Recognizer to Text Spotter (CVPR 2021 · similarity 0.91)
- A Text Detection System for Natural Scenes With Convolutional Feature Learning and Cascaded Classification (CVPR 2016 · similarity 0.90)
- Multi-Oriented Text Detection With Fully Convolutional Networks (CVPR 2016 · similarity 0.90)
- Bridging the Gap Between End-to-End and Two-Step Text Spotting (CVPR 2024 · similarity 0.88)
- OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning (NeurIPS 2025 · 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.