from pdf2image import convert_from_path import os import numpy as np import cv2 from utils import segment_characters, prepare_char_for_model, MODEL_IMAGE_SIZE from config import settings def test_image_processing_utils(): print("Testing image processing utilities with a PDF...") POPPLER_PATH = settings.POPPLER_PATH project_root = os.path.abspath(os.path.join(os.path.dirname(__file__), '')) pdf_name = "Applied-Machine-Learning-and-AI-for-Engineers.pdf" sample_pdf_path = os.path.join(project_root, "sample_documents", pdf_name) try: print(f"Reading first page from '{sample_pdf_path}'...") page_image_pil = convert_from_path( sample_pdf_path, first_page=1, last_page=2, poppler_path=os.path.join(POPPLER_PATH, "bin") )[1] page_image_bgr = cv2.cvtColor(np.array(page_image_pil), cv2.COLOR_RGB2BGR) print("Successfully converted PDF page to image.") gray_image = cv2.cvtColor(page_image_bgr, cv2.COLOR_BGR2GRAY) _, binary_img = cv2.threshold( gray_image, 0, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU ) print(f"Successfully preprocessed image. Shape: {binary_img.shape}") boxes = segment_characters(binary_img) print(f"Found {len(boxes)} potential character bounding boxes.") for x, y, w, h in boxes: cv2.rectangle(page_image_bgr, (x, y), (x + w, y + h), (0, 255, 0), 2) output_path = os.path.join(project_root, "sample_documents", "pdf_segmentation_result.png") cv2.imwrite(output_path, page_image_bgr) print(f"Segmentation visualization saved to: {output_path}") if boxes: x, y, w, h = boxes[0] first_char_crop = binary_img[y:y + h, x:x + w] char_tensor = prepare_char_for_model(first_char_crop) print(f"Prepared first character for model. Tensor shape: {char_tensor.shape}") assert char_tensor.shape == (1, 1, MODEL_IMAGE_SIZE, MODEL_IMAGE_SIZE) print("Tensor shape is correct.") except Exception as e: print(f"An error occurred: {e}") print(f"\nPlease ensure the PDF file exists at the absolute path: '{sample_pdf_path}'") print("Also check that your POPPLER_PATH is correct.") if __name__ == "__main__": test_image_processing_utils()