---
license: cc-by-nc-sa-4.0
language:
  - vi
task_categories:
  - image-to-text
tags:
  - ocr
  - vietnamese
  - han-nom
  - quoc-ngu
  - vintage
pretty_name: vietage-ocr
configs:
  - config_name: default
    data_files:
      - split: train
        path: "train-*.parquet"
      - split: test
        path: "test-*.parquet"
---

# Vietage OCR

Page-level OCR ground truth for vintage Vietnamese texts: Hán-Nôm woodblocks, pre-reform national Vietnamese (Quốc ngữ) books, and 1929–34 newspapers.

## Load

```python
from datasets import load_dataset
ds = load_dataset("taidng/vietage-ocr")
ds["train"][0]  # {"id": ..., "image": PIL, "label": str, "metadata": {...}}
```

## Makeup

| Group         | Train | Test | Source                            |
| --------------- | ----: | ---: | --------------------------------- |
| `wikisource_qn` | 3,200 |  355 | Vietnamese Wikisource             |
| `nomna_nom`     | 2,657 |  295 | NomNaOCR (Kaggle, derived VNPF)   |
| `vnpf_nom`      |   135 |   15 | VNPF (Kiều 1870/1902, CPN)        |
| `pntv_self`     |     0 |   15 | Phụ Nữ Tân Văn newspaper, hand-labeled |
| **Total**       | 5,992 |  680 |                                   |