Research map / Vision-language and multimodal / Human and embodied modeling
TextAdaIN: Paying Attention to Shortcut Learning in Text Recognizers
Oren Nuriel, Sharon Fogel, Ron Litman
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 TextAdaIN | SCATTER · IAM | word accuracy: 85.7 -> 87.3 | Improved |
| add TextAdaIN | SCATTER · RIMES | word accuracy: 93.3 -> 94.4 | Improved |
| add TextAdaIN | SCATTER · CVL | WER: 22.3 -> 21.8 | Improved |
| add TextAdaIN | Baek et al. (CTC) · RIMES | word accuracy: 87.8 -> 90.7 | Improved |
| add TextAdaIN | AbiNet · IAM | word accuracy: 85.4 -> 86.3 | Improved |
| add MixStyle or pAdaIN | SCATTER · IAM | WER: 14.4 -> 14.3 (MixStyle), 14.6 (pAdaIN) | No clear effect |
| operate on channel and height with 5 windows (TextAdaIN) instead of vanilla AdaIN over channels (ablation) | SCATTER · IAM | word accuracy: 85.4 -> 87.3 | Improved |
| add TextAdaIN on top of SCATTER augmentation pipeline (ablation) | SCATTER · IAM | word accuracy: 85.7 -> 87.3 | Improved |
| add TextAdaIN without any augmentation (ablation) | SCATTER · IAM | word accuracy: 84.6 -> 86.5 | Improved |
| use text-image donors instead of Gaussian noise donors (ablation) | TextAdaIN · IAM | word accuracy: 86.1 -> 87.3 | Improved |
| use text-image donors instead of blank-image donors (ablation) | TextAdaIN · IAM | word accuracy: 86.2 -> 87.3 | Improved |
| increase application probability p from 0.01 to 0.1 (ablation) | TextAdaIN · IAM | word accuracy: 87.3 -> 78.0 | Hurt |
Closest papers by meaning
- An End-to-End TextSpotter With Explicit Alignment and Attention (CVPR 2018 · similarity 0.90)
- Multi-modal Text Recognition Networks: Interactive Enhancements between Visual and Semantic Features (ECCV 2022 · similarity 0.89)
- Towards End-To-End Text Spotting With Convolutional Recurrent Neural Networks (ICCV 2017 · similarity 0.88)
- Self-Supervised Learning for Text Recognition: A Critical Survey (IJCV 2025 · similarity 0.88)
- OTE: Exploring Accurate Scene Text Recognition Using One Token (CVPR 2024 · similarity 0.87)
- Bridging the Gap Between End-to-End and Two-Step Text Spotting (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.
