Research map / Architectures and efficiency / CNN design
CNN-N-Gram for Handwriting Word Recognition
Arik Poznanski, Lior Wolf
This paper tried a common change: Add data augmentation
add test-side data augmentation · IAM · WER (%): 6.94 -> 6.45 Improved
What 689 accepted papers report for this change
- 560 improved
- 44 no clear effect
- 26 mixed
- 59 hurt
- Understanding Multimodal Learning: A Loss Landscape Smoothness Perspective ICML 2026 · add data augmentation (jitter, crop, flip) · UPMC-Food101 · accuracy: 92.84 -> 94.93Improved
- AsyncBEV: Cross-modal flow alignment in Asynchronous 3D Object Detection ICLR 2026 · add EMC + asynchronous data augmentation training · nuScenes val, 0 s offset · NDS: 66.7 -> 63.1Hurt
- Towards Understanding Continual Factual Knowledge Acquisition of Language Models: From Theory to Algorithm ICML 2026 · add data augmentation (5 biographies per individual) in pre-training · Biography PT test, unseen templates · EM: 2.71 -> 81.27Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 CNN-predicted n-gram attributes with CCA matching | handwriting word recognition · IAM | WER (%): 11.90 -> 6.45 | Improved |
| use CNN-predicted n-gram attributes with CCA matching | handwriting word recognition · RIMES | WER (%): 4.75 -> 3.90 | Improved |
| use CNN-predicted n-gram attributes with CCA matching | handwriting word recognition · IFN/ENIT abcde-s | accuracy (%): 88.50 -> 94.09 | Improved |
| use CNN-N-Gram | printed word recognition · SVT full | accuracy (%): 80.7 -> 83.62 | Improved |
| use CNN-N-Gram | printed word recognition · IIIT5K-1000 | accuracy (%): 94.4 -> 94.15 | Hurt |
| use probability vectors instead of FC representations as CCA input (ablation) | CNN-N-Gram · IAM | WER (%): 6.45 -> 6.56 | No clear effect |
| remove trigrams during test (ablation) | CNN-N-Gram · IAM | WER (%): 6.45 -> 6.33 | No clear effect |
| remove bigrams and trigrams during test (ablation) | CNN-N-Gram · IAM | WER (%): 6.45 -> 6.29 | No clear effect |
| remove trigrams during training (ablation) | CNN-N-Gram · IAM | WER (%): 6.45 -> 6.32 | No clear effect |
| remove bigrams and trigrams during training (ablation) | CNN-N-Gram · IAM | WER (%): 6.45 -> 6.36 | No clear effect |
| use 19 specialized FC branches rather than 7 merged branches (ablation) | CNN-N-Gram · IAM | WER (%): 7.16 -> 6.45 | Improved |
| use 19 specialized FC branches rather than one FC branch (ablation) | CNN-N-Gram · IAM | WER (%): 7.81 -> 6.45 | Improved |
Closest papers by meaning
- Learning Visual N-Grams From Web Data (ICCV 2017 · similarity 0.82)
- Character n-Gram Embeddings to Improve RNN Language Models (AAAI 2019 · similarity 0.81)
- An Alternative Deep Feature Approach to Line Level Keyword Spotting (CVPR 2019 · similarity 0.81)
- Neural Ctrl-F: Segmentation-Free Query-By-String Word Spotting in Handwritten Manuscript Collections (ICCV 2017 · similarity 0.81)
- N-gram Unsupervised Compoundation and Feature Injection for Better Symbolic Music Understanding (AAAI 2024 · similarity 0.79)
- On the Generalization of Handwritten Text Recognition Models (CVPR 2025 · similarity 0.79)
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.
