feat: Add app

main.py

@@ -1,15 +1,13 @@
 from pathlib import Path
 
 import cv2
-import numpy as np
 import supervision as sv
 import torch
 import yaml
-from pdf2image import convert_from_bytes, convert_from_path
-from PIL import Image
 from tqdm import tqdm
 
 from groundingdino.util.inference import Model, preprocess_caption
+from pdf_converter import PdfConverter
 
 GROUNDING_DINO_CONFIG = "groundingdino/config/GroundingDINO_SwinT_OGC.py"
 GROUNDING_DINO_CHECKPOINT = "gdino_checkpoints/groundingdino_swint_ogc.pth"
@@ -17,49 +15,6 @@ BOX_THRESHOLD = 0.4
 TEXT_THRESHOLD = 0.25
 
 
-class PdfConverterService:
-    """
-    A service to convert PDF files to images and resize images,
-    using pypdfium2 for PDFs and Pillow for images.
-    """
-
-    def __init__(self, dpi: int):
-        self._pdfium_initialized = False  # Track if PDFium needs explicit init/deinit
-        self._dpi = dpi
-
-    def convert_pdf_to_jpg(self, file_path: str) -> list[Image.Image]:
-        """
-        Converts a PDF file to JPG images at different scales.
-        """
-        pil_images = convert_from_path(file_path, dpi=self._dpi)
-        return pil_images
-
-    def resize_image(self, img: Image.Image, size: tuple[int, int]) -> Image.Image:
-        """
-        Resizes a PIL Image to the specified size.
-        """
-        return img.resize(size, Image.LANCZOS)
-
-    @staticmethod
-    def save_image_as_png(img: Image.Image, file_path: str):
-        """
-        Saves a PIL Image as a PNG file.
-        """
-        img.save(file_path, format="PNG")
-
-    @staticmethod
-    def to_cv2_image(img: Image.Image):
-        open_cv_image = np.array(img.convert("RGB"))
-        return open_cv_image[:, :, ::-1].copy()
-
-    def convert_pdf_bytes_to_jpg(self, pdf_bytes: bytes) -> list[Image.Image]:
-        """
-        Converts PDF bytes to JPG images at different scales.
-        """
-        pil_images = convert_from_bytes(pdf_bytes, dpi=self._dpi)
-        return pil_images
-
-
 def main(
     data_dir: str | Path,
     text: str = "ID card. Carte Vitale. Bank details. Human face.",
@@ -82,7 +37,7 @@ def main(
     for img_path in tqdm(
         data_dir.glob("*.pdf"), total=len(list(data_dir.glob("*.pdf")))
     ):
-        pdf_convertor = PdfConverterService(120)
+        pdf_convertor = PdfConverter(120)
         if img_path.suffix == ".pdf":
             imgs = pdf_convertor.convert_pdf_to_jpg(str(img_path))
             img = imgs[0]