Research map / Vision-language and multimodal / Human and embodied modeling
Towards Accurate Scene Text Recognition With Semantic Reasoning Networks
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 |
|---|---|---|---|
| Add transformer units to the backbone (ablation) | SRN · IC13, IC15, IIIT5K, SVT, SVTP, CUTE | accuracy: 90.0 -> 93.0 (IC13); 80.5 -> 83.6 (CUTE) | Helped |
| Add character reading order information to parallel visual attention (ablation) | SRN · IC13, IC15, IIIT5K, SVT, SVTP, CUTE | accuracy: 93.0 -> 93.2 (IC13); 79.8 -> 79.4 (SVTP) | Mixed |
| Add GSRM (ablation) | SRN · IIIT5K and IC15 | accuracy: 92.3 -> 94.8 (IIIT5K); 77.5 -> 82.7 (IC15) | Helped |
| Use 4 rather than 6 transformer units in GSRM (ablation) | SRN · IC15 | accuracy: 82.7 -> 82.7 | No clear effect |
| Use global multi-way semantic reasoning instead of forward one-way reasoning (ablation) | SRN · IC13 | accuracy: 94.7 -> 95.5 | Helped |
| Use global multi-way semantic reasoning instead of backward one-way reasoning (ablation) | SRN · IC13 | accuracy: 94.5 -> 95.5 | Helped |
| Replace gated fusion with addition (ablation) | SRN visual-semantic fusion decoder · IC13 | accuracy: 95.5 -> 95.2 | Hurt |
| Replace gated fusion with concatenation (ablation) | SRN visual-semantic fusion decoder · IC13 | accuracy: 95.5 -> 95.0 | Hurt |
| Replace gated fusion with dot-product fusion (ablation) | SRN visual-semantic fusion decoder · IC13 | accuracy: 95.5 -> 94.8 | Hurt |
| Add GSRM to SRN (ablation) | SRN · TRW-T and TRW-L | character-level accuracy: 80.6 -> 85.5 (TRW-T); 77.5 -> 84.3 (TRW-L) | Helped |
| Use parallel semantic reasoning rather than serial attention decoding | SRN · IC15 and TRW-L | inference time: 191.6ms; 216.8ms | Helped |
Closest papers by meaning
- Visual Semantics Allow for Textual Reasoning Better in Scene Text Recognition (AAAI 2022 · similarity 0.90)
- Image as a Language: Revisiting Scene Text Recognition via Balanced, Unified and Synchronized Vision-Language Reasoning Network (AAAI 2024 · similarity 0.88)
- Scene Text Recognition from Two-Dimensional Perspective (AAAI 2019 · similarity 0.87)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.86)
- Explicit Relational Reasoning Network for Scene Text Detection (AAAI 2025 · similarity 0.85)
- Linguistics-aware Masked Image Modeling for Self-supervised Scene Text Recognition (CVPR 2025 · similarity 0.85)
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.