Research map / Detection and segmentation / CNN features
TextDS: Parameter-Efficient Representation Alignment for Scene Text Detection under Distribution Shifts
Boyuan Chen, Zichen Dang, Chuang Yang, Lap-Pui Chau, Yi Wang
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 Dual-Encoding (SAM2 + DINOv3) (ablation) | TextDS baseline · CTW-1500 | F-measure (%): 82.2 -> 85.3 | Improved |
| add Dual-Encoding (SAM2 + DINOv3) (ablation) | TextDS baseline · Total-Text | F-measure (%): 85.1 -> 86.0 | Improved |
| add Dual-Encoding (SAM2 + DINOv3) (ablation) | TextDS baseline · MLT | F-measure (%): 86.4 -> 88.1 | Improved |
| add SWLoRA to dual encoder (ablation) | TextDS · CTW-1500 | F-measure (%): 85.3 -> 87.9 | Improved |
| add CSF to dual encoder (ablation) | TextDS · CTW-1500 | F-measure (%): 85.3 -> 88.2 | Improved |
| add CSF on top of Dual-Encoding + SWLoRA (ablation) | TextDS · CTW-1500 | F-measure (%): 87.9 -> 90.6 | Improved |
| add SWLoRA to dual encoder (ablation) | TextDS · Total-Text | F-measure (%): 86.0 -> 88.3 | Improved |
| add CSF to dual encoder (ablation) | TextDS · Total-Text | F-measure (%): 86.0 -> 87.8 | Improved |
| add SWLoRA on top of Dual-Encoding + CSF (ablation) | TextDS · MLT | F-measure (%): 91.0 -> 92.2 | Improved |
| early-exit in SWLoRA (max step 5) (ablation) | SWLoRA · CTW-1500 | executed refinement steps: 3.652 out of 5, a 27.0% reduction | Improved |
| early-exit in SWLoRA (max step 5) (ablation) | SWLoRA · Total-Text | executed refinement steps: 2.205 out of 5, a 55.9% reduction | Improved |
| increase LoRA rank or maximum step beyond default (rank 8, step 5) (ablation) | SWLoRA · CTW-1500, Total-Text, MLT | F-measure | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LRANet — LRANet: Towards Accurate and Efficient Scene Text Detection with Low-Rank Approximation Network (AAAI 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MOST: A Multi-Oriented Scene Text Detector With Localization Refinement (CVPR 2021 · similarity 0.88)
- TextNeRF: A Novel Scene-Text Image Synthesis Method based on Neural Radiance Fields (CVPR 2024 · similarity 0.88)
- TextSSR: Diffusion-based Data Synthesis for Scene Text Recognition (ICCV 2025 · similarity 0.88)
- EAST: An Efficient and Accurate Scene Text Detector (CVPR 2017 · similarity 0.88)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.86)
- LRANet: Towards Accurate and Efficient Scene Text Detection with Low-Rank Approximation Network (AAAI 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.
