Research map / Detection and segmentation / CNN features
CentripetalText: An Efficient Text Instance Representation for Scene Text Detection
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 |
|---|---|---|---|
| Relaxed L1 instead of Smooth L1 (ablation) | CentripetalText · Total-Text | F-measure: 83.0 -> 84.9 | Helped |
| Relaxed L1 instead of Smooth L1 (ablation) | CentripetalText · MSRA-TD500 | F-measure: 81.0 -> 83.0 | Helped |
| Relaxed L1 instead of Balanced L1 (ablation) | CentripetalText · Total-Text | F-measure: 83.8 -> 84.9 | Helped |
| CT representation instead of PAN | CentripetalText · MSRA-TD500 | F-measure: 78.9 -> 83.0 | Helped |
| CT representation instead of PAN | CentripetalText · Total-Text | F-measure: 83.5 -> 84.9 | Helped |
| FPEM neck instead of FPN (ablation) | ResNet18 · Total-Text | F-measure: 84.0 -> 84.9 | Helped |
| FPEM neck instead of FPN (speed) (ablation) | ResNet18 · Total-Text | FPS: 46.8 -> 40.0 | Hurt |
| ResNet50 instead of ResNet18 backbone (ablation) | FPEM · Total-Text | F-measure: 84.9 -> 85.6 | Helped |
| pretrain on SynthText | CT-640 · Total-Text | F-measure: 84.9 -> 86.3 | Helped |
| replace SPN with CPN in Mask TextSpotter v3 | Mask TextSpotter v3 · Total-Text end-to-end (no lexicon) | end-to-end F: 71.2 -> 71.9 | Helped |
Closest papers by meaning
- Look More Than Once: An Accurate Detector for Text of Arbitrary Shapes (CVPR 2019 · similarity 0.88)
- Learning Shape-Aware Embedding for Scene Text Detection (CVPR 2019 · similarity 0.88)
- Kernel Adaptive Convolution for Scene Text Detection via Distance Map Prediction (CVPR 2024 · similarity 0.87)
- LRANet: Towards Accurate and Efficient Scene Text Detection with Low-Rank Approximation Network (AAAI 2024 · similarity 0.87)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.86)
- Text Grouping Adapter: Adapting Pre-trained Text Detector for Layout Analysis (CVPR 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.