【处理数据标签】1.提取目标类别的图片和标签；2.删除目标标签制作负样本。

1.提取数据集需要的标签类别的图片和标签另存

VOC格式的xml标签：
（全部代码）

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

def copy_files_with_bottle_label(source_folder, out_folder):
    # 创建输出文件夹
    out_images_folder = os.path.join(out_folder, "images")
    out_xml_folder = os.path.join(out_folder, "xml")
    os.makedirs(out_images_folder, exist_ok=True)
    os.makedirs(out_xml_folder, exist_ok=True)

    # 遍历XML文件夹
    for root, dirs, files in os.walk(os.path.join(source_folder, "xml")):
        for file in files:
            if file.endswith('.xml'):
                xml_path = os.path.join(root, file)
                tree = ET.parse(xml_path)
                root_xml = tree.getroot()
                for obj in root_xml.findall('object'):
                    label = obj.find('name').text
                    if label == 'bottle':
                    # if label in ['bottle', 'littlebottle']:
                        # 复制XML文件到输出文件夹
                        shutil.copy(xml_path, out_xml_folder)
                        # 复制对应的图片文件到输出文件夹
                        img_name = os.path.splitext(file)[0] + '.jpg'
                        img_path = os.path.join(source_folder, "images", img_name)
                        shutil.copy(img_path, out_images_folder)
                        break  # 如果找到了符合条件的标签，就不需要继续检查该XML文件了




source_folders = r"E:\demo\source_folders "
folders_name = ["folder1_202401", "folder2_202402"] #所包含的子文件夹
out_folder = r"E:\demo\out_folder "

for folder_name in folders_name:
    source_folder = os.path.join(source_folders, folder_name)
    # out_folder = os.path.join(out_folder, folder_name)
    copy_files_with_bottle_label(source_folder, out_folder)

2. 删除图片中的目标和标签yolo版，并另存

用白色填充图片的目标区域，删除目标标签类的标签框。
（全部代码）

import os
from PIL import Image, ImageDraw
from tqdm import tqdm

def process_yolo_label(label_path, out_label_path, out_img_path):
    # 打开对应的图片文件
    img_path = os.path.join(os.path.dirname(label_path.replace('labels', 'images')), os.path.splitext(os.path.basename(label_path))[0] + '.jpg')
    img = Image.open(img_path)
    img_width, img_height = img.size

    # 读取标签文件内容
    with open(label_path, 'r') as f:
        lines = f.readlines()

    # 创建新的标签文件内容
    new_lines = []
    for line in lines:
        # 解析标签信息
        parts = line.strip().split()
        label = parts[0]
        x_center = float(parts[1])
        y_center = float(parts[2])
        width = float(parts[3])
        height = float(parts[4])

        # 检查目标类别是否需要保留,保留8,9目标类别。
        # if label in ['8', '9']:
        if label in ['7']:
            new_lines.append(line)
        else:
            # 计算目标区域的坐标
            left = int((x_center - width / 2) * img_width)
            top = int((y_center - height / 2) * img_height)
            right = int((x_center + width / 2) * img_width)
            bottom = int((y_center + height / 2) * img_height)

            # 用白色填充目标区域
            draw = ImageDraw.Draw(img)
            draw.rectangle([(left, top), (right, bottom)], fill='white')

    # 保存修改后的标签文件
    with open(out_label_path, 'w') as f:
        f.writelines(new_lines)

    # 保存修改后的图片文件
    img.save(out_img_path)

def process_folder(source_folder, out_folder):
    out_images_folder = os.path.join(out_folder, "images")
    out_labels_folder = os.path.join(out_folder, "labels")
    os.makedirs(out_images_folder, exist_ok=True)
    os.makedirs(out_labels_folder, exist_ok=True)

    # 获取标签文件夹中的文件数量
    num_files = sum(len(files) for _, _, files in os.walk(os.path.join(source_folder, "labels")))

    # 使用 tqdm 添加进度条
    with tqdm(total=num_files, desc='Processing') as pbar:
        # 遍历标签文件夹
        for root, _, files in os.walk(os.path.join(source_folder, "labels")):
            for file in files:
                if file.endswith('.txt'):
                    label_path = os.path.join(root, file)
                    out_label_path = os.path.join(out_labels_folder, file)
                    out_img_path = os.path.join(out_images_folder, os.path.splitext(file)[0] + '.jpg')
                    process_yolo_label(label_path, out_label_path, out_img_path)
                    pbar.update(1)  # 更新进度条

source_folder = r"E:\demo\source_folder \val"
out_folder = r"E:\demo\out_folder \del_class"
process_folder(source_folder, out_folder)

3. 删除图片中的目标和标签XML版

用白色填充图片的目标区域，删除目标标签类的标签框。
（全部代码）

import os
import shutil
from xml.etree import ElementTree as ET
from PIL import Image, ImageDraw

def process_voc_xml(xml_path, out_xml_path, out_img_path):
    tree = ET.parse(xml_path)
    root = tree.getroot()

    # 创建一份新的XML树
    new_root = ET.Element("annotation")
    for elem in root:
        if elem.tag == 'folder':
            new_folder = ET.SubElement(new_root, elem.tag)
            new_folder.text = elem.text
        elif elem.tag == 'filename':
            new_filename = ET.SubElement(new_root, elem.tag)
            new_filename.text = elem.text
        elif elem.tag == 'size':
            new_size = ET.SubElement(new_root, elem.tag)
            for size_elem in elem:
                new_size_elem = ET.SubElement(new_size, size_elem.tag)
                new_size_elem.text = size_elem.text
        elif elem.tag == 'object':
            name = elem.find('name').text
            if name in ['bottle', 'littlebottle']:
                new_root.append(elem)

    # 保存修改后的XML文件
    new_tree = ET.ElementTree(new_root)
    new_tree.write(out_xml_path)

    # 打开对应的图片文件
    img_name = os.path.splitext(os.path.basename(xml_path))[0] + '.jpg'
    img_path = os.path.join(os.path.dirname(xml_path.replace('xml', 'images')), img_name)
    img = Image.open(img_path)
    draw = ImageDraw.Draw(img)

    # 获取图片的尺寸
    img_width, img_height = img.size

    # 用白色填充图片上除了 "bottle" 和 "littlebottle" 类别之外的目标区域
    for obj in root.findall('object'):
        name = obj.find('name').text
        if name not in ['bottle', 'littlebottle']:
            bndbox = obj.find('bndbox')
            xmin = int(bndbox.find('xmin').text)
            ymin = int(bndbox.find('ymin').text)
            xmax = int(bndbox.find('xmax').text)
            ymax = int(bndbox.find('ymax').text)
            draw.rectangle([(xmin, ymin), (xmax, ymax)], fill='white')

    # 保存修改后的图片文件
    img.save(out_img_path)

def process_folder(source_folder, out_folder):
    out_images_folder = os.path.join(out_folder, "images")
    out_xml_folder = os.path.join(out_folder, "xml")
    os.makedirs(out_images_folder, exist_ok=True)
    os.makedirs(out_xml_folder, exist_ok=True)

    # 遍历XML文件夹
    for root, _, files in os.walk(os.path.join(source_folder, "xml")):
        for file in files:
            if file.endswith('.xml'):
                xml_path = os.path.join(root, file)
                out_xml_path = os.path.join(out_xml_folder, file)
                out_img_path = os.path.join(out_images_folder, os.path.splitext(file)[0] + '.jpg')
                process_voc_xml(xml_path, out_xml_path, out_img_path)

source_folder = r"E:\demo\source_folder "
out_folder = r"E:\demo\source_folder "
process_folder(source_folder, out_folder)