-
Notifications
You must be signed in to change notification settings - Fork 0
/
ijson_parser.py
145 lines (138 loc) · 5.85 KB
/
ijson_parser.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
import json
import ijson
file = '/home/brian/Documents/datasets/openimages/detection_train.json'
def load_images(json_filename):
counter = 0
images = []
annotations = []
categories = []
with open(json_filename, 'rb') as input_file:
# load json iteratively
parser = ijson.parse(input_file)
loading_images = False
loading_annotations = False
loading_categories = False
current_img = {
"id": None,
"width": None,
"height": None,
"file_name": None
}
current_ann = {
"id": None,
"image_id": None,
"area": None,
"iscrowd": None,
"category_id": None,
"bbox": list()
}
current_category = {
"id": None,
"name": None,
}
for prefix, event, value in parser:
if event == 'start_array' and prefix == 'images':
loading_images = True
continue
if event == 'start_array' and prefix == 'annotations':
loading_annotations = True
continue
if event == 'start_array' and prefix == 'categories':
loading_categories = True
continue
if loading_images:
if counter > 100:
counter = 0
loading_images = False
all_are_not_none = all(v is not None for v in current_img.values())
if all_are_not_none:
images.append(current_img)
print('Parsed images={}'.format(len(images)))
current_img = {
"id": None,
"width": None,
"height": None,
"file_name": None
}
counter += 1
else:
if event == 'number' and prefix == 'images.item.id':
current_img['id'] = int(value)
elif event == 'number' and prefix == 'images.item.width':
current_img['width'] = int(value)
elif event == 'number' and prefix == 'images.item.height':
current_img['height'] = int(value)
elif event == 'string' and prefix == 'images.item.file_name':
first_letter = value[0]
current_img['file_name'] = f'train_{first_letter}/{value}'
if prefix == 'images' and event == 'end_array':
loading_images = False
if loading_annotations:
ready = current_ann['id'] is not None and \
current_ann['image_id'] is not None and \
current_ann['area'] is not None and \
current_ann['iscrowd'] is not None and \
current_ann['category_id'] is not None and \
len(current_ann['bbox']) == 4
if ready:
annotations.append(current_ann)
print('Parsed annotations={}'.format(len(annotations)))
current_ann = {
"id": None,
"image_id": None,
"area": None,
"iscrowd": None,
"category_id": None,
"bbox": list()
}
counter += 1
else:
if event == 'number' and prefix == 'annotations.item.id':
current_ann['id'] = int(value)
elif event == 'number' and prefix == 'annotations.item.image_id':
current_ann['image_id'] = int(value)
elif event == 'number' and prefix == 'annotations.item.area':
current_ann['area'] = int(value)
elif event == 'number' and prefix == 'annotations.item.iscrowd':
current_ann['iscrowd'] = int(value)
elif event == 'number' and prefix == 'annotations.item.category_id':
current_ann['category_id'] = int(value)
elif event == 'number' and prefix == 'annotations.item.bbox.item':
current_ann['bbox'].append(int(value))
if prefix == 'annotations' and event == 'end_array':
loading_annotations = False
if counter > 100:
break
if loading_categories:
ready = current_category['id'] is not None and \
current_category['name'] is not None
if ready:
categories.append(current_category)
print('Parsed categories={}'.format(len(categories)))
current_category = {
"id": None,
"name": None
}
else:
if event == 'number' and prefix == 'categories.item.id':
current_category['id'] = int(value)
elif event == 'string' and prefix == 'categories.item.name':
current_category['name'] = value
if prefix == 'categories' and event == 'end_array':
loading_categories = False
return images, annotations, categories
i, a, c = load_images(file)
data = {
'info': {
'year': 2022,
'version': 7,
'description': 'OpenImages datasets',
"url": "https://storage.googleapis.com/openimages/web/index.html",
"date_created": "10-10-2023"
},
'images': i,
'annotations': a,
'categories': c
}
print("Saving Json")
json.dump(data, open('/home/brian/Documents/datasets/openimages/detection_train_curated.json', 'w'))