JSON file to PNG image (batch conversion / image naming / migration / pixel value change) [tips]

  One . Code

Get ready ： Contains many json File folder

Use ： In the code json_folder Place to write json Folder address （‘/’ Keep the symbols ）,png_folder Place to write png Folder address .

effect ：json_folder The folder contains batch generated json Folder , There is less inside label.png picture ,png_folder Folders contain named / Change pixel （ It refers to a single channel grayscale image , The pixel value is only 255 and 0） Of png file .

import os
import shutil
import numpy as np
import cv2
from PIL import Image

#json File expansion （ transformation ）
json_folder = '/'
json_name = os.listdir(json_folder)
os.system("activate labelme")
for i in range(len(json_name)):
    if(os.path.splitext(json_name[i])[1] == ".json"):
        json_path = json_folder + json_name[i]
        os.system("labelme_json_to_dataset " + json_path)

#json To png Batch naming of / Transfer 
png_folder= ''
os.makedirs(png_folder,exist_ok=True)
i = 1
for name in json_name:
	if os.path.isdir(json_folder + name):
		if i < 10:  # 0-9
			newname = "00" + str(i) + '.png'
		elif i > 9 and i < 100:  # 10-100
			newname = "0" + str(i) + '.png'
		elif i > 99 and i < 1000:  # 100-1000
		    newname = str(i) + '.png'
		i += 1
		os.chdir(json_folder + name)
		old_name = 'label.png'
		os.rename(old_name, newname)
		shutil.move(json_folder + name + "/" + newname, png_folder)

# Change pixel value 
png_name = os.listdir(png_folder)
for name in png_name:
	data_source = cv2.imread(png_folder +'/'+ name)
	data = np.array(data_source)
	img_path = png_folder + '/' + name
	for i in range(data[:, :, 0].shape[0]):
		for j in range(data[:, :, 0].shape[1]):
			if data[:, :, 2][i][j]  > 0 :
				data[:, :, 2][i][j] = 255 #Red
				data[:, :, 1][i][j] = 255 #Green
				data[:, :, 0][i][j] = 255 #Blue
	cv2.imwrite(img_path , data)
	png = Image.open(img_path).convert('L').save(img_path)

Two . explain

Preface

When I am doing paper reproduction and applying other data sets to this paper , The dataset lacks the truth map used in segmentation . So use labelme Do a good job in generating truth labels json file , It also needs to be converted into something that the code can use png Format truth map , Some of the functional problems encountered have been integrated （ I'm just a sewing monster of knowledge , Refer to the blog post at the end of the article ）.

1.json File activation

labelme Generated json The file needs to activate the generation folder , Contains four documents .

2. transformation / name / transfer

In four documents png Format pictures are useful to us , So put each json In a folder png Extract to a folder , Complete the conversion .

Because it is a split test task , Use the pixel by pixel comparison of the original image and the truth map , The two figures need to be mapped one by one . Therefore, there are corresponding norms for image naming , Convenient for subsequent code sorted Function to sort , The code adopts the following naming format （eg：000,010,200）.

3. Change the pixel value

After the above operation, a red truth map is obtained , It needs to be converted into a single channel gray image . Here, we use the method of modifying pixel values one by one and then converting them into gray-scale images , It's a little complicated in logic , But it also provides the idea of modifying pixel values in different channels .

4. Check

img_path Write the address of the picture you want to see , Zoom in to see the pixel value , It is found that the grayscale image of a single channel is , Be careful cv2.imread Inside cv2.IMREAD_UNCHANGED To place , Otherwise, the grayscale image will also be viewed as RGB chart .

import cv2
img_path = ''
img = cv2.imread(img_path,cv2.IMREAD_UNCHANGED)
cv2.imshow('photo',img)
key = cv2.waitKey(0)

Reference resources

1. Batch conversion / transfer

use python3 Script labelme Of json Format file to png/xml Format file _pyyniffler The blog of -CSDN Blog _json turn png Script

2. Change pixel value

python- Modify the image size and RGB The pixel values of the three channels （opencv）_ Hula Hula hey blog -CSDN Blog _python Modify the image pixel size
Modify the pixel value of all gray-scale pictures under the folder _ Shensheng's brilliant blog -CSDN Blog