目标检测VOC数据集：用PIL实现xml和图片的同步缩放

import os
from PIL import Image
import platform
import xml.dom.minidom


class HandleXmlImage:

    def get_file_path(self, dir_file):
        self.system_version = platform.platform()
        file_name_list = os.listdir(dir_file)
        for file_name in file_name_list:
            if 'Windows' in self.system_version:
                file_path = dir_file + f'\\{file_name}'
            else:
                file_path = dir_file + f'/{file_name}'
            yield file_path, file_name

    def updage_Image(self, file_path, number_after_change_height, file_name, dir_file):
        img = Image.open(file_path)
        height, width = img.size
        scale = number_after_change_height / height
        height = number_after_change_height
        width = int(width * scale)
        new_img = img.resize((width, height), Image.ANTIALIAS)
        new_path = dir_file + '_new'
        if not os.path.exists(new_path):
            os.makedirs(new_path)
        if 'Windows' in self.system_version:
            new_file = new_path + f'\\{file_name}'
        else:
            new_file = new_path + f'/{file_name}'
        new_img.save(new_file, 'JPEG')
        return scale, height, width

    def updata_xml(self, file_path, scale, height, width, file_name, dir_file):
        self.system_version = 'Window'
        dom = xml.dom.minidom.parse(file_path)
        root = dom.documentElement
        size_object = root.getElementsByTagName("size")[0]
        size_object.getElementsByTagName('height')[0].firstChild.data = height
        size_object.getElementsByTagName('width')[0].firstChild.data = width

        bndbox_object = root.getElementsByTagName("bndbox")[0]
        bndbox_object.getElementsByTagName('xmin')[0].firstChild.data = float(
            bndbox_object.getElementsByTagName('xmin')[0].firstChild.data) * scale
        bndbox_object.getElementsByTagName('ymin')[0].firstChild.data = float(
            bndbox_object.getElementsByTagName('ymin')[0].firstChild.data) * scale
        bndbox_object.getElementsByTagName('xmax')[0].firstChild.data = float(
            bndbox_object.getElementsByTagName('xmax')[0].firstChild.data) * scale
        bndbox_object.getElementsByTagName('ymax')[0].firstChild.data = float(
            bndbox_object.getElementsByTagName('ymax')[0].firstChild.data) * scale

        new_path = dir_file + '_new'
        if not os.path.exists(new_path):
            os.makedirs(new_path)
        if 'Windows' in self.system_version:
            new_file = new_path + f'\\{file_name}'
        else:
            new_file = new_path + f'/{file_name}'

        with open(new_file, 'w', encoding='utf8') as f:
            dom.writexml(f, addindent='  ', encoding='utf8')

    def run(self, xml_path, image_path, number_after_change_height):
        xml_range = self.get_file_path(xml_path)
        image_list = list(self.get_file_path(image_path))

        for index, xml_data in enumerate(xml_range):
            xml_file_path, xml_file_name = xml_data
            image_file_path, image_file_name = image_list[index]
            print(image_file_name, image_file_path)
            scale, height, width = self.updage_Image(image_file_path, number_after_change_height, image_file_name,
                                                     image_path)
            print(scale, height, width)
            self.updata_xml(xml_file_path, scale, height, width, xml_file_name, xml_path)


if __name__ == '__main__':
    xml_path = input('请输入xml文件夹路径>>>:')
    image_path = input('请输入image文件夹路径>>>:')
    number_after_change_height = int(input("请输入改变后高度>>>:"))
    HandleXmlImage().run(xml_path, image_path, number_after_change_height)