将labelme标注的语义分割数据集转换成cityscapes格式（使用记录）

1、labelme标注

图片和json文件保存在同一文件夹下。

2、将灰度值映射到特定的类别，指定颜色

1. 修改label.py文件.

我的路径：E:/WorkSoftware/Install/Anaconda3/envs/labelme/Lib/site-packages/imgviz/

# 在这里修改颜色，修改要对应json_to_dataset.py里面自己的类别，
# 颜色可以自己选择，以下是示例
cmap[1, :] = [0, 255, 0]  # class 1
cmap[2, :] = [255, 0, 0]  # class 2

 以下是完整代码：

def label_colormap(n_label=256, value=None):
    """Label colormap.

    Parameters
    ----------
    n_labels: int
        Number of labels (default: 256).
    value: float or int
        Value scale or value of label color in HSV space.

    Returns
    -------
    cmap: numpy.ndarray, (N, 3), numpy.uint8
        Label id to colormap.

    """
    def bitget(byteval, idx):
        shape = byteval.shape + (8,)
        return np.unpackbits(byteval).reshape(shape)[..., -1 - idx]

    i = np.arange(n_label, dtype=np.uint8)
    r = np.full_like(i, 0)
    g = np.full_like(i, 0)
    b = np.full_like(i, 0)

    i = np.repeat(i[:, None], 8, axis=1)
    i = np.right_shift(i, np.arange(0, 24, 3)).astype(np.uint8)
    j = np.arange(8)[::-1]
    r = np.bitwise_or.reduce(np.left_shift(bitget(i, 0), j), axis=1)
    g = np.bitwise_or.reduce(np.left_shift(bitget(i, 1), j), axis=1)
    b = np.bitwise_or.reduce(np.left_shift(bitget(i, 2), j), axis=1)

    cmap = np.stack((r, g, b), axis=1).astype(np.uint8)

    # 在这里修改颜色，修改要对应json_to_dataset.py里面自己的类别
    cmap[1, :] = [0, 255, 0]  # lane

    if value is not None:
        hsv = color_module.rgb2hsv(cmap.reshape(1, -1, 3))
        if isinstance(value, float):
            hsv[:, 1:, 2] = hsv[:, 1:, 2].astype(float) * value
        else:
            assert isinstance(value, int)
            hsv[:, 1:, 2] = value
        cmap = color_module.hsv2rgb(hsv).reshape(-1, 3)

    return cmap

2. 修改json_to_dataset.py 文件。

我的路径：E:/WorkSoftware/Install/Anaconda3/envs/labelme/Lib/site-packages/labelme/cli/

label_name_to_value = {"_background_": 0}
# 修改为自己的类别，我只有一个类别，所以添加了一个
label_name_to_value = {"_background_": 0, "lane": 1}

3、生成img.png、label.png、label_names.txt、label_viz.png

由于labelme里面的json_to_dataset.py只能转换单张json文件，不能处理文件夹，因此我在自己的项目目录下做了修改，新建文件命名为json_to_dataset_labelme.py。完整代码如下：

# json_to_dataset_labelme.py
import argparse
import base64
import json
import os
import os.path as osp

import imgviz
import PIL.Image
from labelme import utils
from labelme.logger import logger


def convert_json_files(input_folder, output_folder=None):
    for root, dirs, files in os.walk(input_folder):
        for file in files:
            if file.endswith(".json"):
                json_file_path = osp.join(root, file)
                convert_single_json(json_file_path, output_folder)


def convert_single_json(json_file, output_folder=None):
    logger.warning(
        "DEPRECATED: This script will be removed in the near future. "
        "Please use `labelme_export_json` instead."
    )
    logger.warning(
        "NOTE: This script is aimed to demonstrate how to convert a JSON file "
        "to a single image dataset. so it won't handle multiple JSON files to "
        "generate a real-use dataset."
    )

    # 获取 JSON 文件的文件名（不包括扩展名）
    json_file_name = osp.splitext(osp.basename(json_file))[0]

    if output_folder is None:
        out_dir = osp.join(osp.dirname(json_file), json_file_name)
    else:
        out_dir = osp.join(output_folder, json_file_name)
    if not osp.exists(out_dir):
        os.mkdir(out_dir)

    # if output_folder is None:
    #     out_dir = osp.basename(json_file).replace(".", "_")
    #     out_dir = osp.join(osp.dirname(json_file), out_dir)
    # else:
    #     out_dir = output_folder
    # if not osp.exists(out_dir):
    #     os.mkdir(out_dir)

    data = json.load(open(json_file))
    imageData = data.get("imageData")

    if not imageData:
        imagePath = os.path.join(os.path.dirname(json_file), data["imagePath"])
        with open(imagePath, "rb") as f:
            imageData = f.read()
            imageData = base64.b64encode(imageData).decode("utf-8")
    img = utils.img_b64_to_arr(imageData)

    label_name_to_value = {"_background_": 0, "lane": 1}
    for shape in sorted(data["shapes"], key=lambda x: x["label"]):
        label_name = shape["label"]
        if label_name in label_name_to_value:
            label_value = label_name_to_value[label_name]
        else:
            label_value = len(label_name_to_value)
            label_name_to_value[label_name] = label_value
    lbl, _ = utils.shapes_to_label(
        img.shape, data["shapes"], label_name_to_value
    )

    label_names = [None] * (max(label_name_to_value.values()) + 1)
    for name, value in label_name_to_value.items():
        label_names[value] = name

    lbl_viz = imgviz.label2rgb(
        lbl, imgviz.asgray(img), label_names=label_names, loc="rb"
    )

    PIL.Image.fromarray(img).save(osp.join(out_dir, "img.png"))
    utils.lblsave(osp.join(out_dir, "label.png"), lbl)
    PIL.Image.fromarray(lbl_viz).save(osp.join(out_dir, "label_viz.png"))

    with open(osp.join(out_dir, "label_names.txt"), "w") as f:
        for lbl_name in label_names:
            f.write(lbl_name + "\n")

    logger.info(f"Saved to: {out_dir}")


def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("input", help="Input folder containing JSON files or a single JSON file")
    parser.add_argument("-o", "--out", default=None, help="Output folder for converted images")
    args = parser.parse_args()

    input_path = args.input
    output_path = args.out

    if os.path.isdir(input_path):
        convert_json_files(input_path, output_path)
    elif os.path.isfile(input_path):
        convert_single_json(input_path, output_path)
    else:
        logger.error("Invalid input path. Please provide a valid file or folder path.")


if __name__ == "__main__":
    main()

运行命令行，即可：

python json_to_dataset_labelme.py path/your_filename -o path/your_output_filename

 结果如下所示：

4、gtFine标签图像

上面生成的label.png的R通道就是需要的gtFine标签图像。

要确保在提取 R 通道之前标签图（label.png）的 R 通道的像素值应该对应于类别的标签值，而不是原始图像的 R 通道。（步骤2）

# gtFine_labelids.py
import os
import cv2

json_path = './out'
for root, dirs, files in os.walk(json_path):
    for file in files:
        if file.strip() == "label.png":
            file_path = os.path.join(root, file)

            src_image_path = os.path.join(root, 'img.png')

            image = cv2.imread(file_path)
            src = cv2.imread(src_image_path)

            folder_name = os.path.split(root)[-1]
            file_name = folder_name + '_' + file.split('_')[0] + '.png'

            # label处理  直接取R通道的图像就是gtFine
            B, G, R = cv2.split(image)

            path = os.path.join(os.path.abspath('.'), 'gtFine')
            if not os.path.exists(path):
                os.makedirs(path)

            labelId_image = os.path.join(path, file_name.split('.')[0] + '_gtFine_labelIds.' + file_name.split('.')[-1])
            print(labelId_image)
            cv2.imwrite(labelId_image, R)

            src_image_path = os.path.join(os.path.abspath('.'), 'leftImg8bit')
            if not os.path.exists(src_image_path):
                os.makedirs(src_image_path)

            # 保存原始图像
            src_image = os.path.join(src_image_path, file_name)
            cv2.imwrite(src_image, src)

            # 保存原始图像的 R 通道
            # src_image = os.path.join(src_image_path, file_name)
            # cv2.imwrite(src_image, R)

            print(file_name)

5、训练集验证集分割

# train_val_seg.py
import os
import random
import shutil

total_list = []
train_list = []
val_list = []
test_list = []


image_path = './leftImg8bit'
label_path = './gtFine'

# 清空
for dir in ['train', 'val', 'test']:
    image_dir = os.path.join(image_path, dir)
    label_dir = os.path.join(label_path, dir)
    if os.path.exists(image_dir):
        shutil.rmtree(image_dir)
    os.makedirs(image_dir)
    if os.path.exists(label_dir):
        shutil.rmtree(label_dir)
    os.makedirs(label_dir)


for root, dirs, files in os.walk(image_path):
    for file in files:
        if file.endswith('png'):
            total_list.append(file)

total_size = len(total_list)
train_size = int(total_size * 0.7)
val_size = int(total_size * 0.2)

train_list = random.sample(total_list, train_size)
remain_list = list(set(total_list) - set(train_list))
val_list = random.sample(remain_list, val_size)
test_list = list(set(remain_list) - set(val_list))

print(len(total_list))
print(len(train_list))
print(len(val_list))
print(len(test_list))


image_path = './leftImg8bit'
label_path = './gtFine'

# 清空
for dir in ['train', 'val', 'test']:
    image_dir = os.path.join(image_path, dir)
    label_dir = os.path.join(label_path, dir)
    if os.path.exists(image_dir):
        shutil.rmtree(image_dir)
        os.makedirs(image_dir)
    if os.path.exists(label_dir):
        shutil.rmtree(label_dir)
        os.makedirs(label_dir)


for file in total_list:
    image_path_0 = os.path.join(image_path, file)
    label_file = file.split('.')[0] + '_gtFine_labelIds' + '.png'
    label_path_0 = os.path.join(label_path, label_file)
    if file in train_list:
        image_path_1 = os.path.join(image_path, 'train', file)
        shutil.move(image_path_0, image_path_1)

        label_path_1 = os.path.join(label_path, 'train', label_file)
        shutil.move(label_path_0, label_path_1)

    elif file in val_list:
        image_path_1 = os.path.join(image_path, 'val', file)
        shutil.move(image_path_0, image_path_1)

        label_path_1 = os.path.join(label_path, 'val', label_file)
        shutil.move(label_path_0, label_path_1)

    elif file in test_list:
        image_path_1 = os.path.join(image_path, 'test', file)
        shutil.move(image_path_0, image_path_1)

        label_path_1 = os.path.join(label_path, 'test', label_file)
        shutil.move(label_path_0, label_path_1)

参考：

将自己的数据制作成cityscape格式_cityscape 格式label 目录-CSDN博客

 labelme 制作语义分割（Semantic Segmentation）数据，不同labels指定不同颜色-CSDN博客