from paddleocr import PaddleOCR from gliner import GLiNER from PIL import Image import gradio as gr import numpy as np import cv2 import logging import os import tempfile import pandas as pd import io import re import traceback # Configure logging logging.basicConfig(level=logging.INFO) logger = logging.getLogger(__name__) # Set up GLiNER environment variables os.environ['GLINER_HOME'] = './gliner_models' # Load GLiNER model try: logger.info("Loading GLiNER model...") gliner_model = GLiNER.from_pretrained("urchade/gliner_large-v2.1") except Exception as e: logger.error("Failed to load GLiNER model") raise e # Get a random color (used for drawing bounding boxes, if needed) def get_random_color(): return tuple(np.random.randint(0, 256, 3).tolist()) # Fixed line def scan_qr_code(image): try: image_np = np.array(image) qr_detector = cv2.QRCodeDetector() data, _, _ = qr_detector.detectAndDecode(image_np) return data.strip() if data else None except Exception as e: logger.error(f"QR scan failed: {str(e)}") return None def extract_emails(text): email_regex = r"\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b" return re.findall(email_regex, text) def extract_websites(text): website_regex = r"\b(?:https?://)?(?:www\.)?([A-Za-z0-9-]+\.[A-Za-z]{2,})(?:/\S*)?\b" matches = re.findall(website_regex, text) return [m for m in matches if '@' not in m] def clean_phone_number(phone): cleaned = re.sub(r"(?!^\+)[^\d]", "", phone) if len(cleaned) < 9 or (len(cleaned) == 9 and cleaned.startswith("+")): return None return cleaned def normalize_website(url): url = url.lower().replace("www.", "").split('/')[0] if not re.match(r"^[a-z0-9-]+\.[a-z]{2,}$", url): return None return f"www.{url}" def extract_address(ocr_texts): address_keywords = ["block", "street", "ave", "area", "industrial", "road"] address_parts = [] for text in ocr_texts: if any(kw in text.lower() for kw in address_keywords): address_parts.append(text) return " ".join(address_parts) if address_parts else None def inference(img: Image.Image, confidence): try: ocr = PaddleOCR(use_angle_cls=True, lang='en', use_gpu=False, det_model_dir='./models/det/en', cls_model_dir='./models/cls/en', rec_model_dir='./models/rec/en') img_np = np.array(img) result = ocr.ocr(img_np, cls=True)[0] ocr_texts = [line[1][0] for line in result] ocr_text = " ".join(ocr_texts) labels = ["person name", "company name", "job title", "phone number", "email address", "address", "website"] entities = gliner_model.predict_entities(ocr_text, labels, threshold=confidence, flat_ner=True) results = { "Person Name": [], "Company Name": [], "Job Title": [], "Phone Number": [], "Email Address": [], "Address": [], "Website": [], "QR Code": [] } # Process entities with validation for entity in entities: text = entity["text"].strip() label = entity["label"].lower() if label == "phone number": if (cleaned := clean_phone_number(text)): results["Phone Number"].append(cleaned) elif label == "email address" and "@" in text: results["Email Address"].append(text.lower()) elif label == "website": if (normalized := normalize_website(text)): results["Website"].append(normalized) elif label == "address": results["Address"].append(text) elif label == "company name": results["Company Name"].append(text) elif label == "person name": results["Person Name"].append(text) elif label == "job title": results["Job Title"].append(text.title()) # Regex fallbacks results["Email Address"] += extract_emails(ocr_text) results["Website"] += [normalize_website(w) for w in extract_websites(ocr_text)] # Phone number validation seen_phones = set() for phone in results["Phone Number"] + re.findall(r'\+\d{8,}|\d{9,}', ocr_text): if (cleaned := clean_phone_number(phone)) and cleaned not in seen_phones: results["Phone Number"].append(cleaned) seen_phones.add(cleaned) results["Phone Number"] = list(seen_phones) # Address processing if not results["Address"]: if (address := extract_address(ocr_texts)): results["Address"].append(address) # Website normalization seen_websites = set() final_websites = [] for web in results["Website"]: if web and web not in seen_websites: final_websites.append(web) seen_websites.add(web) results["Website"] = final_websites # Company name fallback if not results["Company Name"]: if results["Email Address"]: domain = results["Email Address"][0].split('@')[-1].split('.')[0] results["Company Name"].append(domain.title()) elif results["Website"]: domain = results["Website"][0].split('.')[1] results["Company Name"].append(domain.title()) # Name fallback if not results["Person Name"]: for text in ocr_texts: if re.match(r"^(?:[A-Z][a-z]+\s?){2,}$", text): results["Person Name"].append(text) break # QR Code if (qr_data := scan_qr_code(img)): results["QR Code"].append(qr_data) # Create CSV csv_data = {k: "; ".join(v) for k, v in results.items() if v} with tempfile.NamedTemporaryFile(suffix=".csv", delete=False, mode="w") as tmp_file: pd.DataFrame([csv_data]).to_csv(tmp_file, index=False) csv_path = tmp_file.name return ocr_text, csv_data, csv_path, "" except Exception as e: logger.error(f"Processing failed: {traceback.format_exc()}") return "", {}, None, f"Error: {str(e)}\n{traceback.format_exc()}" # Gradio Interface title = 'Enhanced Business Card Parser' description = 'Accurate entity extraction with combined AI and regex validation' if __name__ == '__main__': demo = gr.Interface( inference, [gr.Image(type='pil', label='Upload Business Card'), gr.Slider(0.1, 1, 0.4, step=0.1, label='Confidence Threshold')], [gr.Textbox(label="OCR Result"), gr.JSON(label="Structured Data"), gr.File(label="Download CSV"), gr.Textbox(label="Error Log")], title=title, description=description, css=".gr-interface {max-width: 800px !important;}" ) demo.launch()