Research map / Detection and segmentation / CNN features
Towards End-to-End Unified Scene Text Detection and Layout Analysis
Check 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 standalone clustering with the integrated end-to-end layout branch | Unified Detector on HierText · HierText test | layout analysis PQ: 52.52 -> 53.60 | Helped |
| Replace the commercial GCP API with Unified Detector | HierText text line detection · HierText test | text line detection PQ: 56.17 -> 62.23 | Helped |
| Add HierText to training data | Ours (word) · ICDAR 17 MLT | word detection F: 76.78 -> 77.24 | Helped |
| Add HierText to training data | Ours (word) · Total-Text | word detection F: 87.94 -> 87.90 | No clear effect |
| Add HierText to training data | Ours (line) · CTW1500 | line detection F: 84.88 -> 85.97 | Helped |
| Add HierText to training data | Ours (line) · MSRA-TD500 | line detection F: 86.69 -> 87.70 | Helped |
| Use line-level rather than word-level detections for clustering (ablation) | Unified Detector · HierText test, 128 object queries | layout analysis PQ: 34.38 -> 51.48 | Helped |
| Increase object queries from 128 to 384 (ablation) | Unified-Detector-Word · HierText test | layout analysis PQ: 34.38 -> 39.11 | Helped |
| Increase object queries from 128 to 384 (ablation) | Unified-Detector-Line · HierText test | layout analysis PQ: 51.48 -> 53.60 | Helped |
| Apply alpha-balancing rather than unbalanced clustering loss (ablation) | Unified Detector · HierText test | layout analysis PQ: 50.04 -> 51.48 | Helped |
| Use focal-style factors instead of alpha-balanced clustering loss (ablation) | Unified Detector · HierText test | layout analysis PQ: 51.48 -> 45.22 | Hurt |
| Add the extra clustering projection layer (ablation) | Unified Detector · HierText test | layout analysis PQ: 48.07 -> 51.48 | Helped |
Closest papers by meaning
- LayoutFormer: Hierarchical Text Detection Towards Scene Text Understanding (CVPR 2024 · similarity 0.91)
- Text Grouping Adapter: Adapting Pre-trained Text Detector for Layout Analysis (CVPR 2024 · similarity 0.89)
- Towards Unified Multi-granularity Text Detection with Interactive Attention (ICML 2024 · similarity 0.88)
- Layout-Agnostic Scene Text Image Synthesis with Diffusion Models (CVPR 2024 · similarity 0.85)
- Kernel Adaptive Convolution for Scene Text Detection via Distance Map Prediction (CVPR 2024 · similarity 0.84)
- LRANet: Towards Accurate and Efficient Scene Text Detection with Low-Rank Approximation Network (AAAI 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.