def _get_bb_of_item()

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

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

    def _get_bb_of_item(self, item, height, width):
        ''' Helper function to find the bounding box (bb) of an item in the xml file.
        All the characters within the item are found and the left-most (min) and right-most (max + length)
        are found. 
        The bounding box emcompasses the left and right most characters in the x and y direction. 

        Parameter
        ---------
        item: xml.etree object for a word/line/form.

        height: int
            Height of the form to calculate percentages of bounding boxes

        width: int
            Width of the form to calculate percentages of bounding boxes

        Returns
        -------
        list
            The bounding box [x, y, w, h] in percentages that encompasses the item.
        '''

        character_list = [a for a in item.iter("cmp")]
        if len(character_list) == 0: # To account for some punctuations that have no words
            return None
        x1 = np.min([int(a.attrib['x']) for a in character_list])
        y1 = np.min([int(a.attrib['y']) for a in character_list])
        x2 = np.max([int(a.attrib['x']) + int(a.attrib['width']) for a in character_list])
        y2 = np.max([int(a.attrib['y']) + int(a.attrib['height'])for a in character_list])

        x1 = float(x1) / width
        x2 = float(x2) / width
        y1 = float(y1) / height
        y2 = float(y2) / height
        bb = [x1, y1, x2 - x1, y2 - y1]
        return bb