README.md

metadata

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 stckmn/ocr-input-Directive017-1761353279 using LightOnOCR, a fast and compact 1B OCR model.

Processing Details

• Source Dataset: stckmn/ocr-input-Directive017-1761353279
• Model: lightonai/LightOnOCR-0.9B-32k-1025
• Vocabulary Size: 32k tokens
• Number of Samples: 21
• Processing Time: 1.2 min
• Processing Date: 2025-10-25 00:50 UTC

Configuration

• Image Column: image
• Output Column: markdown
• Dataset Split: train
• Batch Size: 32
• 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: 80.0%

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

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 LightOnOCR script:

uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/lighton-ocr.py \
    stckmn/ocr-input-Directive017-1761353279 \
    <output-dataset> \
    --vocab-size 32k \
    --image-column image \
    --batch-size 32

Performance

• Processing Speed: ~0.29 images/second
• Benchmark Score: 76.1% overall (across diverse document types)
• Optimization: Native resolution ViT + lightweight decoder

Generated with 🤖 UV Scripts