extract_images/services.py

import pymupdf
from io import BytesIO
from PIL import Image
import pdfplumber
import ast
import google.generativeai as genai
from PIL import Image, ImageDraw
import openai
import requests
import os

# from constants import GEMINI_API_KEY, OPENAI_API_KEY
from utils import (
    draw_boxes,
    pdf_to_images,
    parse_bboxs_gemini_flash,
    convert_pdf_to_images,
    encode_image_to_base64,
)


def extract_images_pymupdf(pdf_file):
    pdf_path = "extract_images/input_docs/uploaded_pdf.pdf"
    with open(pdf_path, "wb") as f:
        f.write(pdf_file)

    doc = pymupdf.open(pdf_path)
    images = []
    for page_idx, page in enumerate(doc):
        for img_index, img in enumerate(doc.get_page_images(page_idx)):
            xref = img[0]
            base_image = doc.extract_image(xref)
            image_bytes = base_image["image"]
            image = Image.open(BytesIO(image_bytes))
            images.append(image)
    return images if images != [] else None


def extract_images_pdfplumber(pdf_file):
    pdf_path = "extract_images/input_docs/uploaded_pdf.pdf"
    with open(pdf_path, "wb") as f:
        f.write(pdf_file)

    images = []
    output_dir = "extract_tables/table_outputs"
    pdf_obj = pdfplumber.open(pdf_path)
    for page_idx, page in enumerate(pdf_obj.pages):
        page_bbox = []
        for image_idx, image in enumerate(page.images):
            page_height = page.height
            image_bbox = (
                image["x0"],
                page_height - image["y1"],
                image["x1"],
                page_height - image["y0"],
            )
            page_bbox.append(image_bbox)
            cropped_page = page.crop(image_bbox)
            image_obj = cropped_page.to_image(resolution=400)
            image_path = os.path.join(
                output_dir, f"image-{page_idx + 1}-{image_idx}.png"
            )
            image_obj.save(image_path)
            image = Image.open(image_path)
            images.append(image)
    return images if images != [] else None


def extract_images_gemini(model, pdf_file):
    gemini_api_key = os.getenv("GEMINI_API_KEY")
    genai.configure(api_key=gemini_api_key)
    gemini_model = genai.GenerativeModel(model)
    prompt = f"Extract the bounding boxes of all the images present in this page. Return the bounding boxes as list of lists. Do not include anyother text or symbols in the output"

    pdf_path = "extract_images/input_docs/uploaded_pdf.pdf"
    with open(pdf_path, "wb") as f:
        f.write(pdf_file)

    images = []
    pdf_images = pdf_to_images(pdf_path)
    for page in pdf_images:
        img = Image.open(page).convert("RGB")
        response = gemini_model.generate_content([img, prompt], stream=False)
        response.resolve()
        print(response.text)

        if model == "gemini-pro-vision":
            page_bbox = ast.literal_eval(response.text)
        elif model == "gemini-1.5-flash-latest":
            page_bbox = parse_bboxs_gemini_flash(response.text)

        image = draw_boxes(page, page_bbox)
        images.append(image)
    return images


def extract_images_gpt(model, pdf_file):
    openai_api_key = os.getenv("OPENAI_API_KEY")
    openai.api_key = openai_api_key
    image_media_type = "image/png"

    pdf_path = "extract_images/input_docs/uploaded_pdf.pdf"
    with open(pdf_path, "wb") as f:
        f.write(pdf_file)

    images = convert_pdf_to_images(pdf_path)
    image_paths = pdf_to_images(pdf_path)

    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {openai.api_key}",
    }

    extracted_images = []
    for page_idx, image in enumerate(images):
        base64_string = encode_image_to_base64(image)
        payload = {
            "model": model,
            "messages": [
                {
                    "role": "user",
                    "content": [
                        {
                            "type": "text",
                            "text": "Extract bounding boxes of all the images present in this page. Return bounding boxes as liat of lists and don't provide any other text in the response.",
                        },
                        {
                            "type": "image_url",
                            "image_url": {
                                "url": f"data:image/jpeg;base64,{base64_string}"
                            },
                        },
                    ],
                }
            ],
        }

        response = requests.post(
            "https://api.openai.com/v1/chat/completions", headers=headers, json=payload
        )
        response_json = response.json()
        print(response_json["choices"][0]["message"]["content"])

        if "choices" in response_json and len(response_json["choices"]) > 0:
            extracted_images.append(
                draw_boxes(
                    image_paths[page_idx],
                    ast.literal_eval(response_json["choices"][0]["message"]["content"]),
                )
            )

    return extracted_images