Research map / Detection and segmentation / Open-set and few-shot detection
Background-Insensitive Scene Text Recognition with Text Semantic Segmentation
Liang Zhao, Zhenyao Wu, Xinyi Wu, Greg Wilsbacher, Song 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 |
|---|---|---|---|
| use pretrained SSN instead of jointed training (no SE) (ablation) | BINet · STR benchmarks original Total | accuracy (%): 91.9 -> 92.3 | Improved |
| use pretrained SSN instead of jointed training (with SE) (ablation) | BINet · STR benchmarks original Total | accuracy (%): 91.3 -> 92.6 | Improved |
| add SE to jointed SSN (ablation) | BINet · STR benchmarks original Total | accuracy (%): 91.9 -> 91.3 | Hurt |
| add SE to jointed SSN (ablation) | BINet · STR benchmarks developed Total* | accuracy (%): 93.2 -> 93.4 | Improved |
| add SE to pretrained SSN (ablation) | BINet · STR benchmarks original Total | accuracy (%): 92.3 -> 92.6 | Improved |
| add SR on top of pretrained SSN + SE (ablation) | BINet · STR benchmarks original Total | accuracy (%): 92.6 -> 93.1 | Improved |
| add SR on top of pretrained SSN + SE (ablation) | BINet · STR benchmarks developed Total* | accuracy (%): 93.8 -> 94.4 | Improved |
| use SE instead of Concat embedding (ablation) | BINet · STR benchmarks original Total | accuracy (%): 85.5 -> 92.6 | Improved |
| use SE instead of Add embedding (ablation) | BINet · STR benchmarks original Total | accuracy (%): 92.3 -> 92.6 | Improved |
| use SE instead of Multiply embedding (ablation) | BINet · STR benchmarks developed Total* | accuracy (%): 93.7 -> 93.8 | Improved |
| train on real data with synthetic initialization (SI+Real') instead of original ABINet MJ+ST training | ABINet · STR benchmarks developed Total* | accuracy (%): 92.6 -> 93.7 | Improved |
| replace ABINet with BINet under SI+Real' training | ABINet · STR benchmarks developed Total* | accuracy (%): 93.7 -> 94.4 | Improved |
Closest papers by meaning
- Revisiting Scene Text Recognition: A Data Perspective (ICCV 2023 · similarity 0.90)
- ST-SAM: Multimodal Scene Text Segmentation with Dense Visual and Sparse Textual Prompts via SAM (AAAI 2026 · similarity 0.89)
- Real-Time Scene Text Detection with Differentiable Binarization (AAAI 2020 · similarity 0.87)
- LISTER: Neighbor Decoding for Length-Insensitive Scene Text Recognition (ICCV 2023 · similarity 0.87)
- CLIP is Almost All You Need: Towards Parameter-Efficient Scene Text Retrieval without OCR (CVPR 2025 · similarity 0.87)
- Boosting Semi-Supervised Scene Text Recognition via Viewing and Summarizing (NeurIPS 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.
