[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6KGSWF2R22Z63PAY5NT6GRY":3,"skill-same:01M6KGSWF2R22Z63PAY5NT6GRY":91},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":26,"publishedAt":45,"linkOnly":46,"markdownContent":47,"markdownBody":48,"frontMatter":49,"refName":70,"commitSha":71,"skillDirectory":72,"forksCount":73,"repositoryHtmlUrl":74,"repositoryHomepageUrl":6,"ownerHtmlUrl":75,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":76,"install":77,"resources":90},"01M6KGSWF2R22Z63PAY5NT6GRY","smart-ocr",null,">","legout\u002Fpi-config","legout","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F1469563?v=4","01M6KGERP7E8TF4XARRKPFJ6HQ","Volker L.","01M6KGERT1FYD049PP1WW923TZ",3,"MIT","https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Finstalled-skills\u002Fsmart-ocr\u002FSKILL.md",[18,23],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000002","文档处理","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",[27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44],"smart ocr","ocr","extract","paddleocr","opus","sonnet","processing","multilingual","zh","en","PDF","text","parsing","office","extraction","gpt","support","MCP","2026-11-10 10:12:30",false,"---\n# ═══════════════════════════════════════════════════════════════════════════════\n# CLAUDE OFFICE SKILL - Enhanced Metadata v2.0\n# ═══════════════════════════════════════════════════════════════════════════════\n\n# Basic Information\nname: smart-ocr\ndescription: \">\"\nversion: \"1.0\"\nauthor: claude-office-skills\nlicense: MIT\n\n# Categorization\ncategory: parsing\ntags:\n  - ocr\n  - extraction\n  - multilingual\n  - paddleocr\ndepartment: All\n\n# AI Model Compatibility\nmodels:\n  recommended:\n    - claude-sonnet-4\n    - claude-opus-4\n  compatible:\n    - claude-3-5-sonnet\n    - gpt-4\n    - gpt-4o\n\n# MCP Tools Integration\nmcp:\n  server: office-mcp\n  tools:\n    - extract_text_from_pdf\n\n# Skill Capabilities\ncapabilities:\n  - ocr_processing\n  - multilingual_support\n\n# Language Support\nlanguages:\n  - en\n  - zh\n---\n\n# Smart OCR Skill\n\n## Overview\n\nThis skill enables intelligent text extraction from images and scanned documents using **PaddleOCR** - a leading OCR engine supporting 100+ languages. Extract text from photos, screenshots, scanned PDFs, and handwritten documents with high accuracy.\n\n## How to Use\n\n1. Provide the image or scanned document\n2. Optionally specify language(s) to detect\n3. I'll extract text with position and confidence data\n\n**Example prompts:**\n- \"Extract all text from this screenshot\"\n- \"OCR this scanned PDF document\"\n- \"Read the text from this business card photo\"\n- \"Extract Chinese and English text from this image\"\n\n## Domain Knowledge\n\n### PaddleOCR Fundamentals\n\n```python\nfrom paddleocr import PaddleOCR\n\n# Initialize OCR engine\nocr = PaddleOCR(use_angle_cls=True, lang='en')\n\n# Run OCR on image\nresult = ocr.ocr('image.png', cls=True)\n\n# Result structure: [[box, (text, confidence)], ...]\nfor line in result[0]:\n    box = line[0]      # [[x1,y1], [x2,y2], [x3,y3], [x4,y4]]\n    text = line[1][0]  # Extracted text\n    conf = line[1][1]  # Confidence score\n    print(f\"{text} ({conf:.2f})\")\n```\n\n### Supported Languages\n\n```python\n# Common language codes\nlanguages = {\n    'en': 'English',\n    'ch': 'Chinese (Simplified)',\n    'cht': 'Chinese (Traditional)',\n    'japan': 'Japanese',\n    'korean': 'Korean',\n    'french': 'French',\n    'german': 'German',\n    'spanish': 'Spanish',\n    'russian': 'Russian',\n    'arabic': 'Arabic',\n    'hindi': 'Hindi',\n    'vi': 'Vietnamese',\n    'th': 'Thai',\n    # ... 100+ languages supported\n}\n\n# Use specific language\nocr = PaddleOCR(lang='ch')  # Chinese\nocr = PaddleOCR(lang='japan')  # Japanese\nocr = PaddleOCR(lang='multilingual')  # Auto-detect\n```\n\n### Configuration Options\n\n```python\nfrom paddleocr import PaddleOCR\n\nocr = PaddleOCR(\n    # Detection settings\n    det_model_dir=None,         # Custom detection model\n    det_limit_side_len=960,     # Max side length for detection\n    det_db_thresh=0.3,          # Binarization threshold\n    det_db_box_thresh=0.5,      # Box score threshold\n    \n    # Recognition settings\n    rec_model_dir=None,         # Custom recognition model\n    rec_char_dict_path=None,    # Custom character dictionary\n    \n    # Angle classification\n    use_angle_cls=True,         # Enable angle classification\n    cls_model_dir=None,         # Custom classification model\n    \n    # Language\n    lang='en',                  # Language code\n    \n    # Performance\n    use_gpu=True,               # Use GPU if available\n    gpu_mem=500,                # GPU memory limit (MB)\n    enable_mkldnn=True,         # CPU optimization\n    \n    # Output\n    show_log=False,             # Suppress logs\n)\n```\n\n### Processing Different Sources\n\n#### Image Files\n```python\n# Single image\nresult = ocr.ocr('image.png')\n\n# Multiple images\nimages = ['img1.png', 'img2.png', 'img3.png']\nfor img in images:\n    result = ocr.ocr(img)\n    process_result(result)\n```\n\n#### PDF Files (Scanned)\n```python\nfrom pdf2image import convert_from_path\n\ndef ocr_pdf(pdf_path):\n    \"\"\"OCR a scanned PDF.\"\"\"\n    # Convert PDF pages to images\n    images = convert_from_path(pdf_path)\n    \n    all_text = []\n    for i, img in enumerate(images):\n        # Save temp image\n        temp_path = f'temp_page_{i}.png'\n        img.save(temp_path)\n        \n        # OCR the image\n        result = ocr.ocr(temp_path)\n        \n        # Extract text\n        page_text = '\\n'.join([line[1][0] for line in result[0]])\n        all_text.append(f\"--- Page {i+1} ---\\n{page_text}\")\n        \n        os.remove(temp_path)\n    \n    return '\\n\\n'.join(all_text)\n```\n\n#### URLs and Bytes\n```python\nimport requests\nfrom io import BytesIO\n\n# From URL\nresponse = requests.get('https:\u002F\u002Fexample.com\u002Fimage.png')\nresult = ocr.ocr(BytesIO(response.content))\n\n# From bytes\nwith open('image.png', 'rb') as f:\n    img_bytes = f.read()\nresult = ocr.ocr(BytesIO(img_bytes))\n```\n\n### Result Processing\n\n```python\ndef process_ocr_result(result):\n    \"\"\"Process OCR result into structured data.\"\"\"\n    \n    lines = []\n    for line in result[0]:\n        box = line[0]\n        text = line[1][0]\n        confidence = line[1][1]\n        \n        # Calculate bounding box\n        x_coords = [p[0] for p in box]\n        y_coords = [p[1] for p in box]\n        \n        lines.append({\n            'text': text,\n            'confidence': confidence,\n            'bbox': {\n                'left': min(x_coords),\n                'top': min(y_coords),\n                'right': max(x_coords),\n                'bottom': max(y_coords),\n            },\n            'raw_box': box\n        })\n    \n    return lines\n\n# Sort by position (top to bottom, left to right)\ndef sort_by_position(lines):\n    return sorted(lines, key=lambda x: (x['bbox']['top'], x['bbox']['left']))\n```\n\n### Text Layout Reconstruction\n\n```python\ndef reconstruct_layout(result, line_threshold=10):\n    \"\"\"Reconstruct text layout from OCR results.\"\"\"\n    \n    lines = process_ocr_result(result)\n    lines = sort_by_position(lines)\n    \n    # Group into logical lines\n    text_lines = []\n    current_line = []\n    current_y = None\n    \n    for line in lines:\n        y = line['bbox']['top']\n        \n        if current_y is None or abs(y - current_y) \u003C line_threshold:\n            current_line.append(line)\n            current_y = y\n        else:\n            # New line\n            text_lines.append(' '.join([l['text'] for l in current_line]))\n            current_line = [line]\n            current_y = y\n    \n    # Add last line\n    if current_line:\n        text_lines.append(' '.join([l['text'] for l in current_line]))\n    \n    return '\\n'.join(text_lines)\n```\n\n## Best Practices\n\n1. **Preprocess Images**: Improve quality before OCR\n2. **Choose Correct Language**: Specify language for better accuracy\n3. **Handle Multi-column**: Process columns separately\n4. **Filter Low Confidence**: Skip results below threshold\n5. **Batch Processing**: Process multiple images efficiently\n\n## Common Patterns\n\n### Image Preprocessing\n```python\nfrom PIL import Image, ImageEnhance, ImageFilter\n\ndef preprocess_image(image_path):\n    \"\"\"Preprocess image for better OCR.\"\"\"\n    img = Image.open(image_path)\n    \n    # Convert to grayscale\n    img = img.convert('L')\n    \n    # Enhance contrast\n    enhancer = ImageEnhance.Contrast(img)\n    img = enhancer.enhance(2.0)\n    \n    # Sharpen\n    img = img.filter(ImageFilter.SHARPEN)\n    \n    # Save preprocessed\n    preprocessed_path = 'preprocessed.png'\n    img.save(preprocessed_path)\n    \n    return preprocessed_path\n```\n\n### Batch OCR with Progress\n```python\nfrom tqdm import tqdm\nfrom concurrent.futures import ThreadPoolExecutor\n\ndef batch_ocr(image_paths, max_workers=4):\n    \"\"\"OCR multiple images in parallel.\"\"\"\n    \n    results = {}\n    \n    def process_single(img_path):\n        result = ocr.ocr(img_path)\n        return img_path, result\n    \n    with ThreadPoolExecutor(max_workers=max_workers) as executor:\n        futures = [executor.submit(process_single, p) for p in image_paths]\n        \n        for future in tqdm(futures, desc=\"Processing OCR\"):\n            path, result = future.result()\n            results[path] = result\n    \n    return results\n```\n\n## Examples\n\n### Example 1: Business Card Reader\n```python\nfrom paddleocr import PaddleOCR\nimport re\n\ndef read_business_card(image_path):\n    \"\"\"Extract contact info from business card.\"\"\"\n    \n    ocr = PaddleOCR(use_angle_cls=True, lang='en')\n    result = ocr.ocr(image_path)\n    \n    # Extract all text\n    all_text = []\n    for line in result[0]:\n        all_text.append(line[1][0])\n    \n    full_text = '\\n'.join(all_text)\n    \n    # Parse contact info\n    contact = {\n        'name': None,\n        'email': None,\n        'phone': None,\n        'company': None,\n        'title': None,\n        'raw_text': full_text\n    }\n    \n    # Email pattern\n    email_match = re.search(r'[\\w\\.-]+@[\\w\\.-]+\\.\\w+', full_text)\n    if email_match:\n        contact['email'] = email_match.group()\n    \n    # Phone pattern\n    phone_match = re.search(r'[\\+\\d][\\d\\s\\-\\(\\)]{8,}', full_text)\n    if phone_match:\n        contact['phone'] = phone_match.group().strip()\n    \n    # Name is usually the largest\u002Ffirst text\n    if all_text:\n        contact['name'] = all_text[0]\n    \n    return contact\n\ncard_info = read_business_card('business_card.jpg')\nprint(f\"Name: {card_info['name']}\")\nprint(f\"Email: {card_info['email']}\")\nprint(f\"Phone: {card_info['phone']}\")\n```\n\n### Example 2: Receipt Scanner\n```python\nfrom paddleocr import PaddleOCR\nimport re\n\ndef scan_receipt(image_path):\n    \"\"\"Extract items and total from receipt.\"\"\"\n    \n    ocr = PaddleOCR(use_angle_cls=True, lang='en')\n    result = ocr.ocr(image_path)\n    \n    lines = []\n    for line in result[0]:\n        text = line[1][0]\n        y_pos = line[0][0][1]\n        lines.append({'text': text, 'y': y_pos})\n    \n    # Sort by vertical position\n    lines.sort(key=lambda x: x['y'])\n    \n    receipt = {\n        'items': [],\n        'subtotal': None,\n        'tax': None,\n        'total': None\n    }\n    \n    for line in lines:\n        text = line['text']\n        \n        # Look for total\n        if 'total' in text.lower():\n            amount = re.search(r'\\$?([\\d,]+\\.?\\d*)', text)\n            if amount:\n                if 'sub' in text.lower():\n                    receipt['subtotal'] = float(amount.group(1).replace(',', ''))\n                else:\n                    receipt['total'] = float(amount.group(1).replace(',', ''))\n        \n        # Look for tax\n        elif 'tax' in text.lower():\n            amount = re.search(r'\\$?([\\d,]+\\.?\\d*)', text)\n            if amount:\n                receipt['tax'] = float(amount.group(1).replace(',', ''))\n        \n        # Look for items (line with price)\n        else:\n            item_match = re.search(r'(.+?)\\s+\\$?([\\d,]+\\.?\\d+)$', text)\n            if item_match:\n                receipt['items'].append({\n                    'name': item_match.group(1).strip(),\n                    'price': float(item_match.group(2).replace(',', ''))\n                })\n    \n    return receipt\n\nreceipt_data = scan_receipt('receipt.jpg')\nprint(f\"Items: {len(receipt_data['items'])}\")\nprint(f\"Total: ${receipt_data['total']}\")\n```\n\n### Example 3: Multi-language Document\n```python\nfrom paddleocr import PaddleOCR\n\ndef ocr_multilingual(image_path, languages=['en', 'ch']):\n    \"\"\"OCR document with multiple languages.\"\"\"\n    \n    all_results = {}\n    \n    for lang in languages:\n        ocr = PaddleOCR(use_angle_cls=True, lang=lang)\n        result = ocr.ocr(image_path)\n        \n        texts = []\n        for line in result[0]:\n            texts.append({\n                'text': line[1][0],\n                'confidence': line[1][1]\n            })\n        \n        all_results[lang] = texts\n    \n    # Merge results, keeping highest confidence\n    merged = {}\n    for lang, texts in all_results.items():\n        for item in texts:\n            text = item['text']\n            conf = item['confidence']\n            \n            if text not in merged or merged[text]['confidence'] \u003C conf:\n                merged[text] = {'confidence': conf, 'language': lang}\n    \n    return merged\n\nresult = ocr_multilingual('bilingual_document.png')\nfor text, info in result.items():\n    print(f\"[{info['language']}] {text} ({info['confidence']:.2f})\")\n```\n\n## Limitations\n\n- Handwritten text accuracy varies\n- Very small text may not be detected\n- Complex backgrounds reduce accuracy\n- Rotated text needs angle classification\n- GPU recommended for best performance\n\n## Installation\n\n```bash\n# CPU version\npip install paddlepaddle paddleocr\n\n# GPU version (CUDA 11.x)\npip install paddlepaddle-gpu paddleocr\n\n# Additional dependencies\npip install pdf2image Pillow\n```\n\n## Resources\n\n- [PaddleOCR GitHub](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleOCR)\n- [Model Zoo](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleOCR\u002Fblob\u002Frelease\u002F2.7\u002Fdoc\u002Fdoc_en\u002Fmodels_list_en.md)\n- [Multi-language Support](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleOCR\u002Fblob\u002Frelease\u002F2.7\u002Fdoc\u002Fdoc_en\u002Fmulti_languages_en.md)\n","# Smart OCR Skill\n\n## Overview\n\nThis skill enables intelligent text extraction from images and scanned documents using **PaddleOCR** - a leading OCR engine supporting 100+ languages. Extract text from photos, screenshots, scanned PDFs, and handwritten documents with high accuracy.\n\n## How to Use\n\n1. Provide the image or scanned document\n2. Optionally specify language(s) to detect\n3. I'll extract text with position and confidence data\n\n**Example prompts:**\n- \"Extract all text from this screenshot\"\n- \"OCR this scanned PDF document\"\n- \"Read the text from this business card photo\"\n- \"Extract Chinese and English text from this image\"\n\n## Domain Knowledge\n\n### PaddleOCR Fundamentals\n\n```python\nfrom paddleocr import PaddleOCR\n\n# Initialize OCR engine\nocr = PaddleOCR(use_angle_cls=True, lang='en')\n\n# Run OCR on image\nresult = ocr.ocr('image.png', cls=True)\n\n# Result structure: [[box, (text, confidence)], ...]\nfor line in result[0]:\n    box = line[0]      # [[x1,y1], [x2,y2], [x3,y3], [x4,y4]]\n    text = line[1][0]  # Extracted text\n    conf = line[1][1]  # Confidence score\n    print(f\"{text} ({conf:.2f})\")\n```\n\n### Supported Languages\n\n```python\n# Common language codes\nlanguages = {\n    'en': 'English',\n    'ch': 'Chinese (Simplified)',\n    'cht': 'Chinese (Traditional)',\n    'japan': 'Japanese',\n    'korean': 'Korean',\n    'french': 'French',\n    'german': 'German',\n    'spanish': 'Spanish',\n    'russian': 'Russian',\n    'arabic': 'Arabic',\n    'hindi': 'Hindi',\n    'vi': 'Vietnamese',\n    'th': 'Thai',\n    # ... 100+ languages supported\n}\n\n# Use specific language\nocr = PaddleOCR(lang='ch')  # Chinese\nocr = PaddleOCR(lang='japan')  # Japanese\nocr = PaddleOCR(lang='multilingual')  # Auto-detect\n```\n\n### Configuration Options\n\n```python\nfrom paddleocr import PaddleOCR\n\nocr = PaddleOCR(\n    # Detection settings\n    det_model_dir=None,         # Custom detection model\n    det_limit_side_len=960,     # Max side length for detection\n    det_db_thresh=0.3,          # Binarization threshold\n    det_db_box_thresh=0.5,      # Box score threshold\n    \n    # Recognition settings\n    rec_model_dir=None,         # Custom recognition model\n    rec_char_dict_path=None,    # Custom character dictionary\n    \n    # Angle classification\n    use_angle_cls=True,         # Enable angle classification\n    cls_model_dir=None,         # Custom classification model\n    \n    # Language\n    lang='en',                  # Language code\n    \n    # Performance\n    use_gpu=True,               # Use GPU if available\n    gpu_mem=500,                # GPU memory limit (MB)\n    enable_mkldnn=True,         # CPU optimization\n    \n    # Output\n    show_log=False,             # Suppress logs\n)\n```\n\n### Processing Different Sources\n\n#### Image Files\n```python\n# Single image\nresult = ocr.ocr('image.png')\n\n# Multiple images\nimages = ['img1.png', 'img2.png', 'img3.png']\nfor img in images:\n    result = ocr.ocr(img)\n    process_result(result)\n```\n\n#### PDF Files (Scanned)\n```python\nfrom pdf2image import convert_from_path\n\ndef ocr_pdf(pdf_path):\n    \"\"\"OCR a scanned PDF.\"\"\"\n    # Convert PDF pages to images\n    images = convert_from_path(pdf_path)\n    \n    all_text = []\n    for i, img in enumerate(images):\n        # Save temp image\n        temp_path = f'temp_page_{i}.png'\n        img.save(temp_path)\n        \n        # OCR the image\n        result = ocr.ocr(temp_path)\n        \n        # Extract text\n        page_text = '\\n'.join([line[1][0] for line in result[0]])\n        all_text.append(f\"--- Page {i+1} ---\\n{page_text}\")\n        \n        os.remove(temp_path)\n    \n    return '\\n\\n'.join(all_text)\n```\n\n#### URLs and Bytes\n```python\nimport requests\nfrom io import BytesIO\n\n# From URL\nresponse = requests.get('https:\u002F\u002Fexample.com\u002Fimage.png')\nresult = ocr.ocr(BytesIO(response.content))\n\n# From bytes\nwith open('image.png', 'rb') as f:\n    img_bytes = f.read()\nresult = ocr.ocr(BytesIO(img_bytes))\n```\n\n### Result Processing\n\n```python\ndef process_ocr_result(result):\n    \"\"\"Process OCR result into structured data.\"\"\"\n    \n    lines = []\n    for line in result[0]:\n        box = line[0]\n        text = line[1][0]\n        confidence = line[1][1]\n        \n        # Calculate bounding box\n        x_coords = [p[0] for p in box]\n        y_coords = [p[1] for p in box]\n        \n        lines.append({\n            'text': text,\n            'confidence': confidence,\n            'bbox': {\n                'left': min(x_coords),\n                'top': min(y_coords),\n                'right': max(x_coords),\n                'bottom': max(y_coords),\n            },\n            'raw_box': box\n        })\n    \n    return lines\n\n# Sort by position (top to bottom, left to right)\ndef sort_by_position(lines):\n    return sorted(lines, key=lambda x: (x['bbox']['top'], x['bbox']['left']))\n```\n\n### Text Layout Reconstruction\n\n```python\ndef reconstruct_layout(result, line_threshold=10):\n    \"\"\"Reconstruct text layout from OCR results.\"\"\"\n    \n    lines = process_ocr_result(result)\n    lines = sort_by_position(lines)\n    \n    # Group into logical lines\n    text_lines = []\n    current_line = []\n    current_y = None\n    \n    for line in lines:\n        y = line['bbox']['top']\n        \n        if current_y is None or abs(y - current_y) \u003C line_threshold:\n            current_line.append(line)\n            current_y = y\n        else:\n            # New line\n            text_lines.append(' '.join([l['text'] for l in current_line]))\n            current_line = [line]\n            current_y = y\n    \n    # Add last line\n    if current_line:\n        text_lines.append(' '.join([l['text'] for l in current_line]))\n    \n    return '\\n'.join(text_lines)\n```\n\n## Best Practices\n\n1. **Preprocess Images**: Improve quality before OCR\n2. **Choose Correct Language**: Specify language for better accuracy\n3. **Handle Multi-column**: Process columns separately\n4. **Filter Low Confidence**: Skip results below threshold\n5. **Batch Processing**: Process multiple images efficiently\n\n## Common Patterns\n\n### Image Preprocessing\n```python\nfrom PIL import Image, ImageEnhance, ImageFilter\n\ndef preprocess_image(image_path):\n    \"\"\"Preprocess image for better OCR.\"\"\"\n    img = Image.open(image_path)\n    \n    # Convert to grayscale\n    img = img.convert('L')\n    \n    # Enhance contrast\n    enhancer = ImageEnhance.Contrast(img)\n    img = enhancer.enhance(2.0)\n    \n    # Sharpen\n    img = img.filter(ImageFilter.SHARPEN)\n    \n    # Save preprocessed\n    preprocessed_path = 'preprocessed.png'\n    img.save(preprocessed_path)\n    \n    return preprocessed_path\n```\n\n### Batch OCR with Progress\n```python\nfrom tqdm import tqdm\nfrom concurrent.futures import ThreadPoolExecutor\n\ndef batch_ocr(image_paths, max_workers=4):\n    \"\"\"OCR multiple images in parallel.\"\"\"\n    \n    results = {}\n    \n    def process_single(img_path):\n        result = ocr.ocr(img_path)\n        return img_path, result\n    \n    with ThreadPoolExecutor(max_workers=max_workers) as executor:\n        futures = [executor.submit(process_single, p) for p in image_paths]\n        \n        for future in tqdm(futures, desc=\"Processing OCR\"):\n            path, result = future.result()\n            results[path] = result\n    \n    return results\n```\n\n## Examples\n\n### Example 1: Business Card Reader\n```python\nfrom paddleocr import PaddleOCR\nimport re\n\ndef read_business_card(image_path):\n    \"\"\"Extract contact info from business card.\"\"\"\n    \n    ocr = PaddleOCR(use_angle_cls=True, lang='en')\n    result = ocr.ocr(image_path)\n    \n    # Extract all text\n    all_text = []\n    for line in result[0]:\n        all_text.append(line[1][0])\n    \n    full_text = '\\n'.join(all_text)\n    \n    # Parse contact info\n    contact = {\n        'name': None,\n        'email': None,\n        'phone': None,\n        'company': None,\n        'title': None,\n        'raw_text': full_text\n    }\n    \n    # Email pattern\n    email_match = re.search(r'[\\w\\.-]+@[\\w\\.-]+\\.\\w+', full_text)\n    if email_match:\n        contact['email'] = email_match.group()\n    \n    # Phone pattern\n    phone_match = re.search(r'[\\+\\d][\\d\\s\\-\\(\\)]{8,}', full_text)\n    if phone_match:\n        contact['phone'] = phone_match.group().strip()\n    \n    # Name is usually the largest\u002Ffirst text\n    if all_text:\n        contact['name'] = all_text[0]\n    \n    return contact\n\ncard_info = read_business_card('business_card.jpg')\nprint(f\"Name: {card_info['name']}\")\nprint(f\"Email: {card_info['email']}\")\nprint(f\"Phone: {card_info['phone']}\")\n```\n\n### Example 2: Receipt Scanner\n```python\nfrom paddleocr import PaddleOCR\nimport re\n\ndef scan_receipt(image_path):\n    \"\"\"Extract items and total from receipt.\"\"\"\n    \n    ocr = PaddleOCR(use_angle_cls=True, lang='en')\n    result = ocr.ocr(image_path)\n    \n    lines = []\n    for line in result[0]:\n        text = line[1][0]\n        y_pos = line[0][0][1]\n        lines.append({'text': text, 'y': y_pos})\n    \n    # Sort by vertical position\n    lines.sort(key=lambda x: x['y'])\n    \n    receipt = {\n        'items': [],\n        'subtotal': None,\n        'tax': None,\n        'total': None\n    }\n    \n    for line in lines:\n        text = line['text']\n        \n        # Look for total\n        if 'total' in text.lower():\n            amount = re.search(r'\\$?([\\d,]+\\.?\\d*)', text)\n            if amount:\n                if 'sub' in text.lower():\n                    receipt['subtotal'] = float(amount.group(1).replace(',', ''))\n                else:\n                    receipt['total'] = float(amount.group(1).replace(',', ''))\n        \n        # Look for tax\n        elif 'tax' in text.lower():\n            amount = re.search(r'\\$?([\\d,]+\\.?\\d*)', text)\n            if amount:\n                receipt['tax'] = float(amount.group(1).replace(',', ''))\n        \n        # Look for items (line with price)\n        else:\n            item_match = re.search(r'(.+?)\\s+\\$?([\\d,]+\\.?\\d+)$', text)\n            if item_match:\n                receipt['items'].append({\n                    'name': item_match.group(1).strip(),\n                    'price': float(item_match.group(2).replace(',', ''))\n                })\n    \n    return receipt\n\nreceipt_data = scan_receipt('receipt.jpg')\nprint(f\"Items: {len(receipt_data['items'])}\")\nprint(f\"Total: ${receipt_data['total']}\")\n```\n\n### Example 3: Multi-language Document\n```python\nfrom paddleocr import PaddleOCR\n\ndef ocr_multilingual(image_path, languages=['en', 'ch']):\n    \"\"\"OCR document with multiple languages.\"\"\"\n    \n    all_results = {}\n    \n    for lang in languages:\n        ocr = PaddleOCR(use_angle_cls=True, lang=lang)\n        result = ocr.ocr(image_path)\n        \n        texts = []\n        for line in result[0]:\n            texts.append({\n                'text': line[1][0],\n                'confidence': line[1][1]\n            })\n        \n        all_results[lang] = texts\n    \n    # Merge results, keeping highest confidence\n    merged = {}\n    for lang, texts in all_results.items():\n        for item in texts:\n            text = item['text']\n            conf = item['confidence']\n            \n            if text not in merged or merged[text]['confidence'] \u003C conf:\n                merged[text] = {'confidence': conf, 'language': lang}\n    \n    return merged\n\nresult = ocr_multilingual('bilingual_document.png')\nfor text, info in result.items():\n    print(f\"[{info['language']}] {text} ({info['confidence']:.2f})\")\n```\n\n## Limitations\n\n- Handwritten text accuracy varies\n- Very small text may not be detected\n- Complex backgrounds reduce accuracy\n- Rotated text needs angle classification\n- GPU recommended for best performance\n\n## Installation\n\n```bash\n# CPU version\npip install paddlepaddle paddleocr\n\n# GPU version (CUDA 11.x)\npip install paddlepaddle-gpu paddleocr\n\n# Additional dependencies\npip install pdf2image Pillow\n```\n\n## Resources\n\n- [PaddleOCR GitHub](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleOCR)\n- [Model Zoo](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleOCR\u002Fblob\u002Frelease\u002F2.7\u002Fdoc\u002Fdoc_en\u002Fmodels_list_en.md)\n- [Multi-language Support](https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddleOCR\u002Fblob\u002Frelease\u002F2.7\u002Fdoc\u002Fdoc_en\u002Fmulti_languages_en.md)\n",{"mcp":50,"name":5,"tags":54,"author":55,"models":56,"license":15,"version":64,"category":39,"languages":65,"department":66,"description":7,"capabilities":67},{"tools":51,"server":53},[52],"extract_text_from_pdf","office-mcp",[28,41,34,30],"claude-office-skills",{"compatible":57,"recommended":61},[58,59,60],"claude-3-5-sonnet","gpt-4","gpt-4o",[62,63],"claude-sonnet-4","claude-opus-4","1.0",[36,35],"All",[68,69],"ocr_processing","multilingual_support","main","3efae68c898153694a2f4aad309c7ff9847b5602","installed-skills\u002Fsmart-ocr",0,"https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config","https:\u002F\u002Fgithub.com\u002Flegout","2026-04-28 16:32:33",{"source":8,"skillName":5,"commands":78,"prompt":88,"zipAvailable":89},[79,82,85],{"tool":80,"command":81},"npx","npx skills add legout\u002Fpi-config --skill smart-ocr",{"tool":83,"command":84},"bunx","bunx skills add legout\u002Fpi-config --skill smart-ocr",{"tool":86,"command":87},"pnpm","pnpm dlx skills add legout\u002Fpi-config --skill smart-ocr","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Finstalled-skills\u002Fsmart-ocr\u002FSKILL.md",true,[],[92,127,139,157,180,201],{"skillNo":93,"name":94,"nameZh":6,"description":95,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":96,"categories":97,"keywords":106,"publishedAt":45},"01M6KGW7MYKZ8MWFKN1V8ZJ2EG","visual-explainer","Generate beautiful, self-contained HTML pages that visually explain systems, code changes, plans, and data. Use when the user asks for a diagram, architecture overview, diff review, plan review, project recap, comparison table, or any visual explanation of technical concepts. Also use proactively when you are about to render a complex ASCII table (4+ rows or 3+ columns) — present it as a styled HTML page instead.","https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Fskills\u002Fvisual-explainer\u002FSKILL.md",[98,101],{"categoryNo":99,"categoryName":100,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000009","前端开发",{"categoryNo":102,"categoryName":103,"taxonomyType":104,"taxonomyTypeDesc":105},"01JBSEEDCAT000000000000033","数据分析师","PROFESSION","适用职业",[107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126],"visual explainer","explainer","visual","files","for","surf","concepts","you","ascii","ascii table","as","present","project","use","view","table","plan","generate","architecture","image",{"skillNo":128,"name":129,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":130,"categories":131,"keywords":134,"publishedAt":45},"01M6KGTMTJR6RXXC8AF29MQQW5","table-extractor","https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Finstalled-skills\u002Ftable-extractor\u002FSKILL.md",[132,133],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[135,122,136,35,36,29,37,44,39,137,42,40,41,32,138,31],"table extractor","camelot","data","tables",{"skillNo":140,"name":141,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":142,"categories":143,"keywords":149,"publishedAt":45},"01M6KGMKKFNC1PRNX2W3E243DR","html-slides","https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Finstalled-skills\u002Fhtml-slides\u002FSKILL.md",[144,147,148],{"categoryNo":145,"categoryName":146,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000003","演示文稿",{"categoryNo":99,"categoryName":100,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[150,151,152,40,31,32,153,35,36,42,154,44,155,156],"html slides","slides","html","revealjs","web","pptx","presentation",{"skillNo":158,"name":159,"nameZh":6,"description":160,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":161,"licenseFileUrl":6,"sourceHtmlUrl":162,"categories":163,"keywords":164,"publishedAt":45},"01M6KGJGVM0TWW632Y6VQQV2WN","using-streamlit-session-state","Using st.session_state to manage state across Streamlit reruns. Use when persisting data, handling widget state, implementing callbacks, or debugging state issues. Covers initialization patterns, widget-state association, and common gotchas.","Apache-2.0","https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Finstalled-skills\u002Fdeveloping-with-streamlit\u002Fskills\u002Fusing-streamlit-session-state\u002FSKILL.md",[],[165,166,167,168,169,170,171,172,173,174,175,176,177,120,178,137,179],"session","state","streamlit","session state","using streamlit session state","debugging","implementing","initialization","Callbacks","widget","across","patterns","manage","common","issues",{"skillNo":181,"name":182,"nameZh":6,"description":183,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":161,"licenseFileUrl":6,"sourceHtmlUrl":184,"categories":185,"keywords":187,"publishedAt":45},"01M6KGJD12GJB7W0ZG66JK00F6","using-streamlit-markdown","Covers all Markdown features in Streamlit including GitHub-flavored syntax plus Streamlit extensions like colored text, badges, Material icons, and LaTeX. Use when formatting text, labels, tooltips, or any text-rendering element.","https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Finstalled-skills\u002Fdeveloping-with-streamlit\u002Fskills\u002Fusing-streamlit-markdown\u002FSKILL.md",[186],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[167,188,189,38,190,191,192,193,194,195,196,120,197,198,199,200],"using streamlit markdown","markdown","labels","features","extensions","rendering","material","element","formatting","syntax","LaTeX","icons","github",{"skillNo":202,"name":203,"nameZh":6,"description":204,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":161,"licenseFileUrl":6,"sourceHtmlUrl":205,"categories":206,"keywords":207,"publishedAt":45},"01M6KGJ7M5Q1RHBZ72MCWTM9GQ","using-streamlit-layouts","Structuring Streamlit app layouts. Use when placing content in sidebars, columns, containers, or dialogs. Covers sidebar usage, column limits, horizontal containers, dialogs, and bordered cards.","https:\u002F\u002Fgithub.com\u002Flegout\u002Fpi-config\u002Fblob\u002Fmain\u002Finstalled-skills\u002Fdeveloping-with-streamlit\u002Fskills\u002Fusing-streamlit-layouts\u002FSKILL.md",[],[167,208,209,210,211,212,120,213,214],"using streamlit layouts","usage","columns","column","app","content","containers"]