TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models
Results and benchmarks
TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models presents a transformer method.
| Task | Dataset | Metric | Value | Source |
|---|---|---|---|---|
| Transformer | IAM Handwriting | CER | 4 | paper-derived |
| Language modeling | From Scratch | Recall | 38.43 | paper-derived |
| Language modeling | + Pretrained Model | Precision | 72.95 | paper-derived |
| Language modeling | + Data Augmentation | Precision | 82.58 | paper-derived |
| Transformer | CRNN | Precision | 48.58 | paper-derived |
| Transformer | Tesseract OCR | Recall | 57.50 | paper-derived |
Audit each benchmark finding before selecting an implementation path. Evidence refs map to the disclosure below.
Evidence graph: 4 refs, 4 links.
Utility signals: depth 100/100, grounding 95/100, status high.
Implementation
Best maintained implementation now
Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities
22,193 stars · 2,702 forks · Last push Jan 23, 2026 · MIT license
- License
- CI
- Dependencies
- Docker
Official implementation from Papers with Code · Community adoption signal (22193 stars)
microsoft/unilm is the strongest maintained implementation based on ranking signals. License is declared (MIT).
Open microsoft/unilm- No CI workflows detected
- Dependency manifest is missing
- Selected microsoft/unilm as the strongest maintained implementation for new work.
- Repository activity is within the last 24 months.
Compare implementation paths
Compare maintenance quality, reproducibility coverage, and evidence confidence before choosing a reproduction baseline.
- Maintenance
- Stale risk
- Confidence
- High
- Reproducibility
- Limited
- Stars
- 22,193
- Last push
- Jan 23, 2026 (215d)
Official implementation from Papers with Code · Community adoption signal (22193 stars)
- No CI pipeline detected
- No Docker setup
- Dependency manifest missing
- Maintenance
- Active
- Confidence
- Low
- Reproducibility
- Strong
- Stars
- 164,399
- Last push
- Aug 24, 2026 (1d)
Community adoption signal (164399 stars)
- No Docker setup
- Low confidence match
- Maintenance
- Stale
- Confidence
- Low
- Reproducibility
- Strong
- Stars
- 758
- Last push
- Aug 22, 2025 (368d)
Community adoption signal (758 stars)
- No push in 12+ months
- No tagged releases
- No Docker setup
Reproduction readiness
Major work
No dependency manifest, manual reconstruction required
- microsoft/unilm has no requirements.txt, environment.yml, pyproject.toml, or Dockerfile.
- You will need to reverse-engineer dependencies from import statements in the source code.
- Last push was 215 days ago.
Hardware requirements
- Expect multi-day setup/compute for meaningful reproduction based on current guidance.
Repositories and ecosystem
No additional verified repositories beyond the primary recommendation.
These repositories had low-confidence matching signals and are hidden by default.
- huggingface/transformers
Confidence: Low · 164,399 stars
- oleehyo/texteller
Confidence: Low · 758 stars
- d-gurgurov/im2latex
Confidence: Low · 20 stars
- DS2BRAIN/ds2
Confidence: Low · 52 stars
- NielsRogge/Vision-Transformer-papers
Confidence: Low · 203 stars
Hugging Face artifacts
No direct paper-linked artifacts were found. Showing strongest curated related artifacts for faster exploration.
Datasets
No trustworthy datasets matches right now.
Search datasets on Hugging FaceSpaces
No trustworthy spaces matches right now.
Search spaces on Hugging FaceResearch context
Tasks
Transformer, Language modeling
Methods
Transformer
Domains
None detected
Open this paper in HFEPX to review benchmark signals, evaluation modes, and human-feedback protocol context.
Open in HFEPXJump to Paper2Code search queries derived from this paper's research context.
Data includes links from Papers with Code ( CC-BY-SA-4.0 ).