labelme标记工具安装及批量转化json文件

直接在anaconda环境安装

安装完之后labelme启动

标记过程省略，标记完之后生成json文件，json文件需要进行转化

批量转化教程

找到labelme安装路径下的json_to_dataset.py

这个自带的json_to_dataset.py一次只能转化一个json，所以要进行批量

更改代码，将以下代码替换原本代码

import argparse
import json
import os
import os.path as osp
import warnings
import PIL.Image
import yaml
from labelme import utils
import base64


def main():
    warnings.warn("This script is aimed to demonstrate how to convert the\n"

                  "JSON file to a single image dataset, and not to handle\n"

                  "multiple JSON files to generate a real-use dataset.")

    parser = argparse.ArgumentParser()

    parser.add_argument('json_file')

    parser.add_argument('-o', '--out', default=None)

    args = parser.parse_args()

    json_file = args.json_file

    if args.out is None:

        out_dir = osp.basename(json_file).replace('.', '_')

        out_dir = osp.join(osp.dirname(json_file), out_dir)

    else:

        out_dir = args.out

    if not osp.exists(out_dir):

        os.mkdir(out_dir)

    count = os.listdir(json_file)

    for i in range(0, len(count)):

        path = os.path.join(json_file, count[i])

        if os.path.isfile(path):

            data = json.load(open(path))
        
            if data['imageData']:

                imageData = data['imageData']

            else:

                imagePath = os.path.join(os.path.dirname(path), 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}

            for shape in data['shapes']:

                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

            # label_values must be dense

            label_values, label_names = [], []

            for ln, lv in sorted(label_name_to_value.items(), key=lambda x: x[1]):

                label_values.append(lv)

                label_names.append(ln)

            assert label_values == list(range(len(label_values)))
 
            lbl = utils.shapes_to_label(img.shape, data['shapes'], label_name_to_value)

            captions = ['{}: {}'.format(lv, ln)

                for ln, lv in label_name_to_value.items()]

            lbl_viz = utils.draw_label(lbl, img, captions)
           
            out_dir = osp.basename(count[i]).replace('.', '_')

            out_dir = osp.join(osp.dirname(count[i]), out_dir)

            if not osp.exists(out_dir):

                os.mkdir(out_dir)

            PIL.Image.fromarray(img).save(osp.join(out_dir, 'img.png'))

            #PIL.Image.fromarray(lbl).save(osp.join(out_dir, 'label.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')

            warnings.warn('info.yaml is being replaced by label_names.txt')

            info = dict(label_names=label_names)

            with open(osp.join(out_dir, 'info.yaml'), 'w') as f:

                yaml.safe_dump(info, f, default_flow_style=False)

            print('Saved to: %s' % out_dir)

                  
if __name__ == '__main__':

    main()

cd 路径这是你转化之后保存文件的路径

labelme_json_to_dataset C:\Users\Admini\Desktop\111（这个路径是存放json的路径）

注：如果报错module 'labelme.utils' has no attribute 'draw_label'，是因为新版本(比如4.5版本)的做了更改，可以替换，用旧版本pip install labelme==3.6