按XML标签的名称，自动给图片分类到对应文件夹，并统计和输出 个数，且还可以恢复原状

功能

1. 分类并统计

把图片分类到对应类比文件夹（根据XML的标签名字自动生成）

统计各类别个数，输出，并以txt格式保存至VOC2007根目录下

  * 图片多标签时，图片会被分到multi_targets文件夹下

2. 单类合并

把图片文件还回到 JPEGImages，并删除category文件夹

使用方法

改root：根目录的绝对路径

1. 分类并统计：

用 # 注释掉 def main 函数的 back_to_one_folder 那一行

2. 单类合并：​

​取消注释。

【不要】动 classify_and_count 那一行

右键，运行（run）

代码

*此代码遇到错误直接跳过，个人实际应用，可视具体情况自行添加。
如果标签都对，则不用改

import os
import shutil
import xml.etree.ElementTree as ET
import argparse


root = "path/to/root"


def main(root):
    # 分类并统计
    category_folder, images_folder = classify_and_count(root)

    # 合并
    #back_to_one_folder(category_folder, images_folder)


def classify_and_count(root):
    labels_folder = os.path.join(root, 'Annotations')
    images_folder = os.path.join(root, 'JPEGImages')
    output_folder = os.path.join(images_folder, 'category')
    os.makedirs(output_folder, exist_ok=True)

    multi_targets_lines = []  # 用于存储multi_targets的内容
    class_counts = {}  # 用于存储类别统计的数量
    multi_num = 0

    for label_file in os.listdir(labels_folder):
        if label_file.endswith(".xml"):
            try:
                tree = ET.parse(os.path.join(labels_folder, label_file))
                tree_root = tree.getroot()
                label_name_elements = tree_root.findall('.//name')

                if not label_name_elements:
                    continue

                img = os.path.splitext(label_file)[0] + '.jpg'
                source_path = os.path.join(images_folder, img)

                if len(label_name_elements) > 1:
                    multi_num += 1
                    class_folder = os.path.join(output_folder, 'multi_targets')
                    for element in label_name_elements:
                        multi_targets_lines.append("{:25s}: {:>25}".format(img, element.text))
                    multi_targets_lines.append("\n")
                else:
                    label_name = label_name_elements[0].text
                    class_folder = os.path.join(output_folder, label_name)
                    class_counts[label_name] = class_counts.get(label_name, 0) + 1

                os.makedirs(class_folder, exist_ok=True)
                shutil.move(source_path, os.path.join(class_folder, img))

            except ET.ParseError:
                pass
            except FileNotFoundError:
                pass

    print("Images have been moved to the corresponding folders.")

    count_txt = os.path.join(root, 'classes_num.txt')
    with open(count_txt, "w") as f:
        for label, count in class_counts.items():
            print("{:25s}: {:>5}个".format(label, count))
            f.write("{:25s}: {:>5}个\n".format(label, count))

        if multi_targets_lines:
            print("{:25s}: {:>5}个\n".format("multi_targets", multi_num))
            f.write("{:25s}: {:>5}个\n".format("multi_targets", multi_num))
            f.write("\n\n以下为多标签图片：\n")
            for line in multi_targets_lines:
                f.write(line + "\n")

    print(f"File {count_txt} has been created and written with content.")
    return output_folder, images_folder


def back_to_one_folder(category_folder, images_folder):
    # 遍历根文件夹下的所有文件（穿透子文件）
    for category_path, _, files in os.walk(category_folder):  # os.walk 返回 (当前文件夹路径, 当前文件夹的子文件夹列表, 当前文件夹中的文件列表)
        for file in files:
            if file.endswith('.jpg'):
                source_file_path = os.path.join(category_path, file)
                destination_file_path = os.path.join(images_folder, file)
                if not os.path.exists(destination_file_path):
                    shutil.move(source_file_path, destination_file_path)
                    print(f"已移动文件: {source_file_path} 到 {destination_file_path}")
                else:
                    print(f"文件已存在，跳过: {source_file_path}")

    shutil.rmtree(category_folder)
    print("All classified images have been moved back.")


if __name__ == '__main__':
    # 创建 ArgumentParser 对象
    parser = argparse.ArgumentParser(description='Description of your script.')
    # 添加命令行参数
    parser.add_argument('--root', type=str, default=root, help='Root path for processing.')
    # 解析命令行参数
    args = parser.parse_args()

    # 调用主函数，并传入解析后的参数
    main(args.root)

运行结果

终端

文件夹

图片

根目录：分类统计，txt