以VOC2007数据集为例，如果我们只想要车辆行人检测，那么只需要从其中选取有行人，自行车，车辆等目标的图片和标签，于是直接上代码：

 

import os
import shutil
import xml.etree.ElementTree as ETdef extract_objects(image_folder, xml_folder, output_image_folder, output_xml_folder, classes):# 创建输出图片文件夹和输出XML文件夹if not os.path.exists(output_image_folder):os.makedirs(output_image_folder)if not os.path.exists(output_xml_folder):os.makedirs(output_xml_folder)# 遍历xml文件夹中的每个XML文件for xml_file in os.listdir(xml_folder):print("Processing:", xml_file)xml_path = os.path.join(xml_folder, xml_file)# 解析XML文件tree = ET.parse(xml_path)root = tree.getroot()# 检查每个object标签，如果其中包含目标类别，则复制对应的图片和XML文件到输出文件夹found_objects = Falsefor obj in root.findall('object'):obj_class = obj.find('name').text.lower()if obj_class in classes:found_objects = Trueprint("Found object:", obj_class)image_name = os.path.splitext(xml_file)[0] + '.jpg'  # 假设图片文件和XML文件名一致，后缀不同image_path = os.path.join(image_folder, image_name)output_image_path = os.path.join(output_image_folder, image_name)output_xml_path = os.path.join(output_xml_folder, xml_file)# 打印路径以进行调试print("Copying from:", image_path)print("Copying to:", output_image_path)print("Copying to:", output_xml_path)# 复制图片和XML文件到输出文件夹try:shutil.copy(image_path, output_image_path)shutil.copy(xml_path, output_xml_path)except Exception as e:print("Error copying files:", str(e))break  # 只要一个目标被提取，即可跳出内循环if not found_objects:print("No target objects found in:", xml_file)# 设置输入文件夹路径和输出文件夹路径
image_folder = 'VOCdevkit/VOC2007/JPEGImages/'
xml_folder = 'VOCdevkit/VOC2007/Annotations/'
output_image_folder = 'VOCdevkit/voc/JPEGImages/'
output_xml_folder = 'VOCdevkit/voc/Annotations/'# 指定需要提取的类别
classes = ['person', 'bicycle', 'bus', 'car', 'motorbike']# 执行提取操作
extract_objects(image_folder, xml_folder, output_image_folder, output_xml_folder, classes)