def _process_data()

in ocr/utils/iam_dataset.py [0:0]

• 35 lines of code
• 7 McCabe index (conditional complexity)

    def _process_data(self):
        ''' Function that iterates through the downloaded xml file to gather the input images and the
        corresponding output.
        
        Returns
        -------
        pd.DataFrame
            A pandas dataframe that contains the subject, image and output requested.
        '''
        image_data = []
        xml_files = glob.glob(self._root + "/xml/*.xml")
        print("Processing data:")
        logging.info("Processing data")
        for i, xml_file in enumerate(xml_files):
            tree = ET.parse(xml_file)
            root = tree.getroot()
            height, width = int(root.attrib["height"]), int(root.attrib["width"])
            for item in root.iter(self._parse_method.split("_")[0]):
                # Split _ to account for only taking the base "form", "line", "word" that is available in the IAM dataset
                if self._parse_method in ["form", "form_bb", "form_original"]:
                    image_id = item.attrib["id"]
                else:
                    tmp_id = item.attrib["id"]
                    tmp_id_split = tmp_id.split("-")
                    image_id = os.path.join(tmp_id_split[0], tmp_id_split[0] + "-" + tmp_id_split[1], tmp_id)
                image_filename = os.path.join(self._root, self._parse_method.split("_")[0], image_id + ".png")
                image_arr = self._pre_process_image(image_filename)
                if image_arr is None:
                    continue
                output_data = self._get_output_data(item, height, width)
                if self._parse_method == "form_bb":
                    image_arr, output_data = self._crop_and_resize_form_bb(item, image_arr, output_data, height, width)
                image_data.append([item.attrib["id"], image_arr, output_data])
                self._reporthook(i, 1, len(xml_files))
        image_data = pd.DataFrame(image_data, columns=["subject", "image", "output"])
        self._save_dataframe_chunks(image_data, self.image_data_file_name)
        return image_data