converts the json file downloaded using image classifer tool of dataturks to dataset folder

tensorflow_json_parser.py

#This script has been solely created under dataturks. Copyrights are reserved

#EXAMPLE USAGE
#python3 tensorflow_json_parser.py --json_file "flower.json" --dataset_path "Dataset5/" 



import json 
import glob
import urllib.request
import argparse
import random
import os
from pathlib import Path


def downloader(image_url , i):
    file_name = str(i)
    full_file_name = str(file_name) + '.jpg'
    prefix = 'http://'
    offset=7
    if image_url[:5] == 'https':
        prefix = 'https://'
        offset=8
    escaped_url = prefix + urllib.request.quote(image_url[offset:])
    print(escaped_url)
    urllib.request.urlretrieve(escaped_url,full_file_name)

if __name__ == "__main__":
	a = argparse.ArgumentParser()
	a.add_argument("--json_file", help="path to json")
	a.add_argument("--dataset_path", help="path to the dataset")
	args = a.parse_args()

	if args.json_file is None and args.dataset_path is None:
	  a.print_help()
	  sys.exit(1)


	with open(args.json_file) as file1:
		lis = []	
		for i in file1:
			lis.append(json.loads(i))

	folder_names = []
	label_to_urls = {}

	for i in lis:
		if len(i['annotation']['labels']) == 0:
			continue
		if i['annotation']['labels'][0] not in folder_names:
			folder_names.append(i['annotation']['labels'][0])
			label_to_urls[i['annotation']['labels'][0]] = [i['content']] 	
		else:
			label_to_urls[i['annotation']['labels'][0]].append(i['content'])

	print(label_to_urls.keys())

	Path(args.dataset_path).mkdir(parents=True, exist_ok=True)
	os.chdir(args.dataset_path)
	for i in label_to_urls.keys():
		Path(str(i)).mkdir(parents=True, exist_ok=True)
		os.chdir(str(i))
		k = 0;	
		for j in label_to_urls[i]:
			b = os.path.basename(j)
			if not Path.exists(Path(b)):
				downloader(j , b)
			k+=1
		os.chdir("../")

skip empty label sets

escape urls as needed

detect protocol

only download if not exists

keep original filename