Research map / Vision-language and multimodal / Human and embodied modeling
Measuring Human Perception to Improve Handwritten Document Transcription
Samuel Grieggs, Bingyu Shen, Greta Rauch, Pei Li, Jiaqi Ma, David Chiang, Brian L. Price, Walter J. Scheirer
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 psychophysical loss (easy priority) to CTC | CRNN · RIMES test | CER (%): 7.47 -> 6.33 | Improved |
| add psychophysical loss (easy priority) | CRNN · IAM test | CER (%): 8.00 -> 7.76 | Improved |
| add psychophysical loss (easy priority) | CTransformer · Latin test | CER (%): 9.81 -> 8.96 | Improved |
| add psychophysical loss (easy priority) | URNN · RIMES test | CER (%): 6.75 -> 6.66 | No clear effect |
| add psychophysical loss (easy priority) | CTransformer · Latin validation | WER (%): 54.93 -> 55.39 | Hurt |
| prioritize hard examples instead of easy (ablation) | psychophysical loss · IAM test (CRNN) | CER (%): 7.76 -> 7.56 | Improved |
| prioritize hard examples instead of easy (ablation) | psychophysical loss · CTransformer all datasets | CER | Hurt |
| replace reaction times with algorithmic hard example mining penalties (ablation) | psychophysical loss · IAM test (CRNN) | CER (%): 8.00 -> 7.76 | No clear effect |
| use reaction times for curriculum learning instead of loss (ablation) | CRNN · IAM test | CER (%): 8.00 -> 7.68 | Improved |
| use text length as penalty instead of reaction time (ablation) | psychophysical loss · IAM test (CRNN) | CER (%): 8.00 -> 7.92 | No clear effect |
| add psychophysical loss (H) with language model and synthetic pretraining | CRNN · IAM test | CER (%): 4.71 -> 4.21 | Improved |
| reduce image height from 128 to 64 (ablation) | CRNN · speed | images per second: 16 -> 31 | Mixed |
Closest papers by meaning
- Seeing is Believing? Mitigating OCR Hallucinations in Multimodal Large Language Models (NeurIPS 2025 · similarity 0.83)
- Measuring Per-Unit Interpretability at Scale Without Humans (NeurIPS 2024 · similarity 0.81)
- Reference-Based Post-OCR Processing with LLM for Precise Diacritic Text in Historical Document Recognition (AAAI 2025 · similarity 0.81)
- ASCIIEval: Benchmarking Models' Visual Perception in Text Strings via ASCII Art (ICLR 2026 · similarity 0.81)
- HRVDA: High-Resolution Visual Document Assistant (CVPR 2024 · similarity 0.81)
- 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.
