Through the built-in code in the labelme environment, convert json to png format in batches.
Table of contents
1. Install the labelme environment
2. Modify the json_to_dataset file
1. Install the labelme environment
How to install and use ylabelme
Install the labelme environment through the method in the above link.
2. Modify the json_to_dataset file
Find the json_to_dataset file in the labelme environment.
Refer to my file location: "D:\Software\Anaconda3\envs\labelme\Lib\site-packages\labelme\cli\json_to_dataset.py"
After opening, directly replace the original code with the following code
'''
修改后的json_to_dataset文件，直接复制替换你自己原始的json_to_dataset
'''
import argparse
import base64
import json
import os
import os.path as osp
import imgviz
import PIL.Image
from labelme.logger import logger
from labelme import utils
def main():
logger.warning(
"This script is aimed to demonstrate how to convert the "
"JSON file to a single image dataset."
)
logger.warning(
"It won't handle 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
print(osp.dirname(json_file))
if osp.isdir(osp.join(osp.dirname(json_file), 'json_data')) is False:
os.mkdir(osp.join(osp.dirname(json_file), 'json_data'))
else:
print("文件已存在")
if args.out is None:
out_dir = osp.basename(json_file).replace(".", "_")
out_dir1 = osp.join(osp.dirname(json_file), 'json_data')
out_dir = osp.join(out_dir1, out_dir)
print(out_dir)
print("#" * 10)
else:
out_dir = args.out
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}
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("Saved to: {}".format(out_dir))
if __name__ == "__main__":
main()3. Batch conversion
Create a new .py file
'''
将label中标注的json文件，转化为可用于分割训练的标签二值化黑白png图片
'''
import os
import cv2
import numpy as np
import shutil
import glob
# def json_png(): 第一次转换用到
path_json = r'D:\Study\SeaIce\sea_ice_data\Data\GF-3\B\3\ERDAS\hv_Elee_cj\hv_2_Elee_cj_0-255_512\lable_2\chonghua' # 这里是指.json文件所在文件夹的路径
# 批量转换，修改此路径
# 此路径为，json文件所在路径
# def extract_png(): 第二次转换用到
path_json_to_data = os.path.join(path_json, "json_data") # json文件夹所在位置
path_save_png = os.path.join(path_json, "json_png") # 将标签图从json文件中批量取出后指定保存的文件目录
path_save_png_binary = os.path.join(path_json, "json_png_binary") # 二至图像最终保存的路径
def pre_treatment():
'''
创建三个文件夹用于存储
json_data用于存储json转换img.png label.png label_names.txt label_viz.png的文件夹
json_png用于存储从json_data提取出来的label。png（最终存储名字与json文件对应）
json_png_binary 用于存储最终转换后的8位的单通道黑白图像
:return:
'''
if os.path.isdir(os.path.join(path_json, "json_data")) is False:
os.mkdir(os.path.join(path_json, "json_data"))
else:
print('文件已存在')
if os.path.isdir(os.path.join(path_json, "json_png")) is False:
os.mkdir(os.path.join(path_json, "json_png"))
else:
print('文件已存在')
if os.path.isdir(os.path.join(path_json, "json_png_binary")) is False:
os.mkdir(os.path.join(path_json, "json_png_binary"))
else:
print('文件已存在')
def json_png():
'''
批量将json转换为img.png label.png label_names.txt label_viz.png
并存储至当前文件夹下的json_date文件夹中
:return: 无
'''
json_file = glob.glob(os.path.join(path_json, "*.json"))
os.system("activate labelme") # 激活labelme环境（根据自己设置的修改）
for file in json_file:
os.system("labelme_json_to_dataset.exe %s" % (file)) # 调用labelme，自带的程序进行批量转换
# labelme中\.conda\envs\labelme\Lib\site-packages\labelme\cli中的json_to_dataset.py被修改过
# 具体修改见json_to_dataset.py
def extract_png():
'''
将标签图从json文件中批量取出
:return:
'''
for eachfile in os.listdir(path_json_to_data):
path1 = os.path.join(path_json_to_data, eachfile) # 获取单个json文件夹的目录
if os.path.isdir(path1): # 判断path1路径是否存在
if os.path.exists(path1 + '/label.png'): # 判断path1路径下label.png是否存在
path1 = os.path.join(path1, 'label.png') # 获取PNG所在的路径，准备等待复制
path2 = os.path.join(path_save_png, (eachfile.split('_')[0] + '.png')) # 将png复制到path2路径下的文件夹中去
shutil.copy(path1, path2) # 将path1文件复制到path2
print(eachfile + ' successfully moved')
def png_to_binary():
'''
由于数据集是做二分类分割，所以，需要将ground_truth转换为8位的单通道黑白图像，才能作为训练时的label使用。
将提取出来的png转换为8位的单通道黑白图像
'''
for im in os.listdir(path_save_png):
img = cv2.imread(os.path.join(path_save_png, im))
b, g, r = cv2.split(img)
r[np.where(r != 0)] = 255
cv2.imwrite(os.path.join(path_save_png_binary, im), r)
def process():
pre_treatment() # 预处理，创建存储所需的相应文件夹
json_png() # 调用labelme的json转换png程序
extract_png() # 从转换的数据中提取png图像
png_to_binary() # 将png转换为8位的单通道黑白图像，用于分割训练
if __name__ == "__main__":
process()Run in the labelme environment created in the first step.
Complete the conversion.
Note: The cv2 package needs to be installed in the environment
pip install opencv-python