Document OCR using dots.ocr

This dataset contains OCR results from images in biglam/bpl-card-catalog using DoTS.ocr, a compact 1.7B multilingual model.

Processing Details

• Source Dataset: biglam/bpl-card-catalog
• Model: rednote-hilab/dots.ocr
• Number of Samples: 256
• Processing Time: 2.3 min
• Processing Date: 2025-10-07 14:32 UTC

Configuration

• Image Column: image
• Output Column: markdown
• Dataset Split: train
• Batch Size: 128
• Prompt Mode: layout-all
• Max Model Length: 8,192 tokens
• Max Output Tokens: 8,192
• GPU Memory Utilization: 80.0%

Model Information

DoTS.ocr is a compact multilingual document parsing model that excels at:

• 🌍 100+ Languages - Multilingual document support
• 📊 Table extraction - Structured data recognition
• 📐 Formulas - Mathematical notation preservation
• 📝 Layout-aware - Reading order and structure preservation
• ⚡ Fast inference - 2-3x faster than native HF with vLLM
• 🎯 Compact - Only 1.7B parameters

Dataset Structure

The dataset contains all original columns plus:

• markdown: The extracted text in markdown format
• 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 DoTS OCR script:

uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/dots-ocr.py \
    biglam/bpl-card-catalog \
    <output-dataset> \
    --image-column image \
    --batch-size 128 \
    --prompt-mode layout-all \
    --max-model-len 8192 \
    --max-tokens 8192 \
    --gpu-memory-utilization 0.8

Performance

• Processing Speed: ~1.9 images/second
• GPU Configuration: vLLM with 80% GPU memory utilization

Generated with 🤖 UV Scripts