Research map / Architectures and efficiency / RNNs and normalization
Joint Line Segmentation and Transcription for End-to-End Handwritten Paragraph Recognition
Read the paperCheck 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 a BLSTM decoder after standard collapse (ablation) | MDLSTM-RNN on IAM · IAM validation, line-level training | CER%: 8.4 -> 7.5 | Helped |
| Add a BLSTM decoder after standard collapse (ablation) | MDLSTM-RNN on Rimes · Rimes validation, line-level training | CER%: 4.9 -> 4.8 | Helped |
| Replace standard collapse with attention-based weighted collapse (ablation) | MDLSTM-RNN with BLSTM decoder on IAM · IAM validation | CER%: 7.5 -> 6.8 | Helped |
| Replace standard collapse with attention-based weighted collapse (ablation) | MDLSTM-RNN with BLSTM decoder on Rimes · Rimes validation | CER%: 4.8 -> 2.5 | Helped |
| Use an attention-based collapse rather than ground-truth line segmentation | MDLSTM-RNN on IAM · IAM validation, 150 dpi | CER%: 8.4 -> 6.8 | Helped |
| Use an attention-based collapse rather than ground-truth line segmentation | MDLSTM-RNN on Rimes · Rimes validation, 300 dpi | CER%: 3.6 -> 2.5 | Helped |
| Add a language model | Weighted collapse on IAM, 150 dpi · IAM test | CER%: 10.1 -> 6.5 | Helped |
| Increase input resolution from 150 dpi to 300 dpi | Weighted collapse on IAM with language model · IAM test | WER%: 16.6 -> 16.4 | No clear effect |
Closest papers by meaning
- OrigamiNet: Weakly-Supervised, Segmentation-Free, One-Step, Full Page Text Recognition by learning to unfold (CVPR 2020 · similarity 0.84)
- Recursive Recurrent Nets With Attention Modeling for OCR in the Wild (CVPR 2016 · similarity 0.83)
- MetaWriter: Personalized Handwritten Text Recognition Using Meta-Learned Prompt Tuning (CVPR 2025 · similarity 0.83)
- General Detection-based Text Line Recognition (NeurIPS 2024 · similarity 0.83)
- MetaHTR: Towards Writer-Adaptive Handwritten Text Recognition (CVPR 2021 · similarity 0.83)
- On the Generalization of Handwritten Text Recognition Models (CVPR 2025 · similarity 0.81)
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.