model_utils.py

import cv2
import pytesseract
import re
from PIL import Image
from ultralytics import YOLO

# Path to your trained YOLO model
MODEL_PATH = "yolov8m_invoiceOCR.pt"

# YOLO class names (order matters)
class_names = [
    "Discount_Percentage", "Due_Date", "Email_Client", "Name_Client", "Products",
    "Remise", "Subtotal", "Tax", "Tax_Precentage", "Tel_Client", "billing address",
    "header", "invoice date", "invoice number", "shipping address", "total"
]

# Load YOLOv8 model
model = YOLO(MODEL_PATH)

def initialize_data_dict():
    return {label: [] if label == "Products" else "" for label in class_names}

def parse_products(raw_text):
    structured = []
    lines = raw_text.split('\n')
    for line in lines:
        match = re.match(r"(\d+)\s+(.*)\s+([\d,]+\.\d{2})\s+([\d,]+\.\d{2})", line)
        if match:
            qty, desc, unit_price, amount = match.groups()
            structured.append({
                "qty": qty,
                "description": desc.strip(),
                "unit_price": unit_price,
                "amount": amount
            })
        elif line.strip():
            structured.append({
                "qty": 0,
                "description": line.strip(),
                "unit_price": 0,
                "amount": 0
            })
    return structured

def extract_invoice_data_from_image(image_path: str):
    image_bgr = cv2.imread(image_path)
    image_rgb = cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB)
    pil_img = Image.fromarray(image_rgb)

    results = model(image_path)[0]
    data = initialize_data_dict()

    for box in results.boxes:
        x1, y1, x2, y2 = map(int, box.xyxy[0])
        cls_id = int(box.cls[0])
        label = class_names[cls_id]

        cropped_img = pil_img.crop((x1, y1, x2, y2))
        extracted_text = pytesseract.image_to_string(cropped_img, config='--psm 6').strip()

        if label == "Products" and extracted_text:
            structured_products = parse_products(extracted_text)
            data["Products"].extend(structured_products)
        elif extracted_text:
            data[label] = extracted_text

    return data