Research map / Detection and segmentation / Open-set and few-shot detection
Turning a CLIP Model Into a Scene Text Spotter
Wenwen Yu, Yuliang Liu, Xingkui Zhu, Haoyu Cao, Xing Sun, Xiang Bai
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 |
|---|---|---|---|
| replace ResNet50 with plain CLIP ResNet50 backbone | DBNet · IC15 detection | F-measure: 87.3 -> 87.7 | No clear effect |
| plain CLIP backbone on CTW1500 | DBNet · CTW1500 detection | F-measure: 83.4 -> 83.4 | No clear effect |
| use FastTCM-CR50 backbone instead of ResNet50 | DBNet · MSRA-TD500 detection | F-measure: 84.9 -> 88.9 | Improved |
| use FastTCM-CR50 instead of TCM-CR50 | DBNet · MSRA-TD500 detection | F-measure: 88.8 -> 88.9 | Improved |
| use FastTCM-CR50 backbone instead of ResNet50 | ABCNet · Total-Text end-to-end | H-mean: 64.2 -> 69.5 | Improved |
| use FastTCM-CR50 backbone instead of ResNet50 | DeepSolo · Total-Text end-to-end | H-mean: 79.7 -> 79.9 | No clear effect |
| use FastTCM-CR50 instead of ResNet50 with few training data | ABCNet · Total-Text few-shot 10% data | H-mean: 43.5 -> 52.3 | Improved |
| use FastTCM-CR50 on SynthText to IC15 adaptation | DBNet · SynthText to IC15 adaptation | F-measure: 64.0 -> 77.2 | Improved |
| use FastTCM-CR50 on TT to IC15 spotting adaptation | ABCNet · TT to IC15 | H-mean: 32.5 -> 48.2 | Improved |
| add predefined language prompt 'Text' (ablation) | DBNet FastTCM-CR50 · MSRA-TD500 detection | F-measure: 87.4 -> 88.9 | Improved |
| add bimodal similarity matching (ablation) | TESTR FastTCM-CR50 · TT to IC15 adaptation | H-mean: 48.2 -> 53.1 | Improved |
| add language prompt generator (ablation) | TESTR FastTCM-CR50 · TT to IC15 adaptation | H-mean: 45.5 -> 53.1 | Improved |
Closest papers by meaning
- Diffusion Feedback Helps CLIP See Better (ICLR 2025 · similarity 0.87)
- Demystifying CLIP Data (ICLR 2024 · similarity 0.87)
- CLIP is Almost All You Need: Towards Parameter-Efficient Scene Text Retrieval without OCR (CVPR 2025 · similarity 0.87)
- GoMatching: A Simple Baseline for Video Text Spotting via Long and Short Term Matching (NeurIPS 2024 · similarity 0.86)
- SuperCLIP: CLIP with Simple Classification Supervision (NeurIPS 2025 · similarity 0.86)
- Building Vision-Language Models on Solid Foundations with Masked Distillation (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.
