ai-training/convert-annotations.py

import os
import shutil
import xml.etree.cElementTree as et
from pathlib import Path

labels_dir = 'labels'
images_dir = 'images'

tag_size = 'size'
tag_object = 'object'
tag_name = 'name'
tag_bndbox = 'bndbox'
# 1 Вантажівка, 2 Машина легкова
name_class_map = {'Truck': 1, 'Car': 2, 'Taxi': 2}


def convert_xml(folder):
    for f in os.listdir(folder):
        if not f.endswith('.jpg'):
            continue

        os.makedirs(images_dir, exist_ok=True)
        os.makedirs(labels_dir, exist_ok=True)

        shutil.copy(os.path.join(folder, f), os.path.join(images_dir, f))
        label = f'{Path(f).stem}.xml'

        tree = et.parse(os.path.join(folder, label))
        root = tree.getroot()
        lines = []
        size_dict = {size_ch.tag : size_ch.text for size_ch in root.findall(f'{tag_size}/*')}
        width = int(size_dict['width'])
        height = int(size_dict['height'])
        for node_object in tree.findall(tag_object):
            for node_object_ch in node_object:
                if node_object_ch.tag == tag_name:
                    class_num = name_class_map[node_object_ch.text]
                if node_object_ch.tag == tag_bndbox:
                    bbox_dict = {bbox_ch.tag: bbox_ch.text for bbox_ch in node_object_ch}
                    xmin = int(bbox_dict['xmin'])
                    xmax = int(bbox_dict['xmax'])
                    ymin = int(bbox_dict['ymin'])
                    ymax = int(bbox_dict['ymax'])
                    c_w = (xmax - xmin) / width
                    c_h = (ymax - ymin) / height
                    c_x = xmin / width + c_w / 2
                    c_y = ymin / height + c_h / 2
            lines.append(f'{class_num} {c_x} {c_y} {c_w} {c_h}')

        with open(os.path.join(labels_dir, f'{Path(label).stem}.txt'), 'w') as f:
            f.writelines(lines)
            f.close()


if __name__ == '__main__':
    convert_xml('datasets/others/UAVimages')