---
tags:
- ocr
- document-processing
- lighton-ocr
- markdown
- uv-script
- generated
---

# Document OCR using LightOnOCR-0.9B-32k-1025

This dataset contains OCR results from images in [NationalLibraryOfScotland/Britain-and-UK-Handbooks-Dataset](https://huggingface.co/datasets/NationalLibraryOfScotland/Britain-and-UK-Handbooks-Dataset) using LightOnOCR, a fast and compact 1B OCR model.

## Processing Details

- **Source Dataset**: [NationalLibraryOfScotland/Britain-and-UK-Handbooks-Dataset](https://huggingface.co/datasets/NationalLibraryOfScotland/Britain-and-UK-Handbooks-Dataset)
- **Model**: [lightonai/LightOnOCR-0.9B-32k-1025](https://huggingface.co/lightonai/LightOnOCR-0.9B-32k-1025)
- **Vocabulary Size**: 32k tokens
- **Number of Samples**: 4,096
- **Processing Time**: 27.8 min
- **Processing Date**: 2025-10-24 12:35 UTC

### Configuration

- **Image Column**: `image`
- **Output Column**: `markdown`
- **Dataset Split**: `train`
- **Batch Size**: 4096
- **Target Image Size**: 1288px (longest dimension)
- **Max Model Length**: 8,192 tokens
- **Max Output Tokens**: 6,500
- **Temperature**: 0.2
- **Top P**: 0.9
- **GPU Memory Utilization**: 93.0%

## Model Information

LightOnOCR is a fast, compact OCR model that excels at:
- ⚡ **Production Speed** - 5.71 pages/second on H100 GPU
- 🎯 **Compact Size** - Only 1B parameters
- 📐 **LaTeX formulas** - Mathematical notation in LaTeX format
- 📊 **Tables** - Extracted and formatted as markdown
- 📝 **Document structure** - Hierarchy and layout preservation
- 🌍 **Multilingual** - Optimized for European languages
- 🔤 **Flexible vocabulary** - 151k/32k/16k token variants

### Vocabulary Variants

- **151k tokens**: Full vocabulary, supports all languages
- **32k tokens**: European languages optimized (~12% faster decoding)
- **16k tokens**: European languages optimized (~12% faster decoding)

## Dataset Structure

The dataset contains all original columns plus:
- `markdown`: The extracted text in markdown format with LaTeX formulas
- `inference_info`: JSON list tracking all OCR models applied to this dataset

## Usage

```python
from datasets import load_dataset
import json

# Load the dataset
dataset = load_dataset("{output_dataset_id}", split="train")

# Access the markdown text
for example in dataset:
    print(example["markdown"])
    break

# View all OCR models applied to this dataset
inference_info = json.loads(dataset[0]["inference_info"])
for info in inference_info:
    print(f"Column: {info['column_name']} - Model: {info['model_id']}")
```

## Reproduction

This dataset was generated using the [uv-scripts/ocr](https://huggingface.co/datasets/uv-scripts/ocr) LightOnOCR script:

```bash
uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/lighton-ocr.py \
    NationalLibraryOfScotland/Britain-and-UK-Handbooks-Dataset \
    <output-dataset> \
    --vocab-size 32k \
    --image-column image \
    --batch-size 4096
```

## Performance

- **Processing Speed**: ~2.46 images/second
- **Benchmark Score**: 76.1% overall (across diverse document types)
- **Optimization**: Native resolution ViT + lightweight decoder

Generated with 🤖 [UV Scripts](https://huggingface.co/uv-scripts)