OCR.py

import os
import torch
from transformers import AutoProcessor, PaliGemmaForConditionalGeneration
from PIL import Image
import io
import re
HF_TOKEN = os.environ.get("HF_TOKEN")
os.environ['PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION'] = 'python'

device = "cuda:0" if torch.cuda.is_available() else "cpu"
model_id = "google/paligemma-3b-mix-224"
model = PaliGemmaForConditionalGeneration.from_pretrained(model_id).to(device)
processor = AutoProcessor.from_pretrained(model_id)

def extract_text_from_image(image_content):
    image = Image.open(io.BytesIO(image_content))
    prompt =  "Extract the following details from this invoice: Invoice Number, Total Amount, Invoice Date."
    inputs = processor(text=prompt, images=image, return_tensors="pt").to(device)
    input_len = inputs["input_ids"].shape[-1]
    with torch.inference_mode():
        generation = model.generate(**inputs, max_new_tokens=100, do_sample=False)
        generation = generation[0][input_len:]
        decoded = processor.decode(generation, skip_special_tokens=True)
    return decoded


def extract_invoice_details(text):
    details = {}
    details['Invoice Number'] = re.search(r'Invoice Number: (\S+)', text).group(1) if re.search(r'Invoice Number: (\S+)', text) else 'N/A'
    details['Amount'] = re.search(r'Total Amount Due: (\S+)', text).group(1) if re.search(r'Total Amount Due: (\S+)', text) else 'N/A'
    details['Invoice Date'] = re.search(r'Invoice Date: (\S+)', text).group(1) if re.search(r'Invoice Date: (\S+)', text) else 'N/A'
    return details