README.md

metadata

tags:
  - ocr
  - text-extraction
  - rolmocr
  - uv-script
  - generated

OCR Text Extraction using RolmOCR

This dataset contains extracted text from images in davanstrien/playbills-pdf-images-text using RolmOCR.

Processing Details

• Source Dataset: davanstrien/playbills-pdf-images-text
• Model: reducto/RolmOCR
• Number of Samples: 10
• Processing Time: 5.8 minutes
• Processing Date: 2025-08-04 17:08 UTC

Configuration

• Image Column: image
• Output Column: rolmocr_text
• Dataset Split: train
• Batch Size: 16
• Max Model Length: 24,000 tokens
• Max Output Tokens: 16,000
• GPU Memory Utilization: 80.0%

Model Information

RolmOCR is a fast, general-purpose OCR model based on Qwen2.5-VL-7B architecture. It extracts plain text from document images with high accuracy and efficiency.

Dataset Structure

The dataset contains all original columns plus:

• rolmocr_text: The extracted text from each image
• 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 extracted text
for example in dataset:
    print(example["rolmocr_text"])
    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 RolmOCR script:

uv run https://huggingface.co/datasets/uv-scripts/ocr/raw/main/rolm-ocr.py \
    davanstrien/playbills-pdf-images-text \
    <output-dataset> \
    --image-column image \
    --batch-size 16 \
    --max-model-len 24000 \
    --max-tokens 16000 \
    --gpu-memory-utilization 0.8

Performance

• Processing Speed: ~0.0 images/second
• GPU Configuration: vLLM with 80% GPU memory utilization

Generated with 🤖 UV Scripts