openocr-skill

Example 1: Batch OCR with Progress

from openocr import OpenOCR
import os
def batch_ocr(image_dir, output_dir='./ocr_results'):
    """OCR all images in a directory."""
    ocr = OpenOCR(task='ocr', mode='mobile')
    os.makedirs(output_dir, exist_ok=True)
    image_files = [
        f for f in os.listdir(image_dir)
        if f.lower().endswith(('.jpg', '.jpeg', '.png', '.bmp', '.tiff'))
    ]
    all_results = {}
    for i, filename in enumerate(image_files):
        filepath = os.path.join(image_dir, filename)
        print(f"[{i+1}/{len(image_files)}] Processing: {filename}")
        results, time_dicts = ocr(
            image_path=filepath,
            save_dir=output_dir,
            is_visualize=True,
        )
        texts = []
        for result in results:
            for line in result:
                texts.append(line[1][0])
        all_results[filename] = texts
        print(f"  Found {len(texts)} text lines")
    # Save all text
    with open(os.path.join(output_dir, 'all_text.txt'), 'w') as f:
        for filename, texts in all_results.items():
            f.write(f"--- {filename} ---\n")
            f.write('\n'.join(texts))
            f.write('\n\n')
    return all_results
results = batch_ocr('./images')

Example 2: Document to Markdown Converter

from openocr import OpenOCR
import os
def doc_to_markdown(input_path, output_dir='./markdown_output'):
    """Convert document images or PDFs to Markdown."""
    doc = OpenOCR(
        task='doc',
        use_layout_detection=True,
        use_chart_recognition=True,
    )
    os.makedirs(output_dir, exist_ok=True)
    result = doc(image_path=input_path)
    if isinstance(result, list):
        # PDF: multiple pages
        for page_result in result:
            doc.save_to_markdown(page_result, output_dir)
        print(f"Converted {len(result)} pages to Markdown")
    else:
        # Single image
        doc.save_to_markdown(result, output_dir)
        print("Converted image to Markdown")
    print(f"Output saved to: {output_dir}")
Convert a scanned PDF
doc_to_markdown('paper.pdf')
Convert a document image
doc_to_markdown('page.jpg')

Example 3: Multi-Task Comparison

from openocr import OpenOCR
def compare_tasks(image_path):
    """Compare results from different OpenOCR tasks."""
    # 1. Detection only
    det = OpenOCR(task='det')
    det_result = det(image_path=image_path)
    num_boxes = len(det_result[0]['boxes'])
    print(f"Detection: Found {num_boxes} text regions")
    # 2. End-to-End OCR
    ocr = OpenOCR(task='ocr')
    ocr_results, _ = ocr(image_path=image_path)
    ocr_texts = [line[1][0] for result in ocr_results for line in result]
    print(f"OCR: Extracted {len(ocr_texts)} text lines")
    for t in ocr_texts[:5]:
        print(f"  - {t}")
    # 3. Universal Recognition
    unirec = OpenOCR(task='unirec')
    text, _ = unirec(image_path=image_path)
    print(f"UniRec: {text[:200]}...")
    return {
        'det_boxes': num_boxes,
        'ocr_texts': ocr_texts,
        'unirec_text': text,
    }
compare_tasks('document.jpg')

Example 4: Gradio Demo Launch

from openocr import launch_openocr_demo, launch_unirec_demo, launch_opendoc_demo
Launch OCR demo
launch_openocr_demo(share=True, server_port=7860, server_name='0.0.0.0')
Launch UniRec demo
launch_unirec_demo(share=True, server_port=7861)
Launch OpenDoc demo
launch_opendoc_demo(share=True, server_port=7862)

1. Choose the Right Task: Use ocr for general text, unirec for formulas/tables, doc for full documents 2. Use Mobile Mode for Speed: mode='mobile' is much faster; use mode='server' only when accuracy is critical 3. Use ONNX Backend: Default ONNX backend works on CPU without extra dependencies 4. Set Appropriate Thresholds: Adjust drop_score (OCR) and layout_threshold (Doc) for your use case 5. Enable Layout Detection: For documents with mixed content (text + formulas + tables), always enable use_layout_detection 6. Batch Processing: Use rec_batch_num to control recognition batch size for throughput optimization 7. GPU Acceleration: Install onnxruntime-gpu or PyTorch with CUDA for significant speedup