-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcombine_datasets.py
226 lines (190 loc) · 8.61 KB
/
combine_datasets.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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
# -*- coding: utf-8 -*-
import cv2
from glob import glob
from tqdm import tqdm
import json
import matplotlib.pyplot as plt
# In[]: MIPT DATASET
#folder_mipt = '/home/kenny/dgx/home/datasets/ir/mipt/'
folder_mipt = '/home/datasets/ir/mipt/'
# In[]:
with open(folder_mipt + 'train.json') as json_file:
train_data = json.load(json_file)
annotations_mipt = train_data['annotations']
categories_mipt = train_data['categories']
images_mipt = train_data['images']
# In[]:
for ann in annotations_mipt:
ann['bbox'] = list(map(int, ann['bbox']))
x, y, w, h = ann['bbox']
ann['segmentation'] = [[x, y, x, y+h, x+w, y+h, x+w, y]]
for img in images_mipt:
img['file_name'] = folder_mipt + 'images/' + img['name']
del(img['name'])
objects_count = len(annotations_mipt)
images_count = len(images_mipt)
#train_data_mipt = {'annotations': annotations_mipt,
# 'categories': categories_mipt,
# 'images': images_mipt,
# 'info': {'contributor': 'no contributor specified',
# 'date_created': 'today',
# 'description': '',
# 'url': 'no url specified',
# 'version': 1.0,
# 'year': 2019},
# 'licenses': []
# }
#with open('train_data_mipt.json', 'w') as outfile:
# json.dump(train_data_mipt, outfile)
# In[]:
#i = 0
#for ann in tqdm(annotations_mipt):
#
## ann = annotations_mipt[0]
# img = cv2.imread(images_mipt[ann['image_id']]['file_name'], 0)
# bbox = ann['bbox']
# cv2.rectangle(img, (int(bbox[0]), int(bbox[1])), (int(bbox[0]+bbox[2]), int(bbox[1]+bbox[3])), (255, 0, 0), 2)
# cv2.putText(img, categories_mipt[ann['category_id']-1]['name'], (50,50), cv2.FONT_HERSHEY_SIMPLEX, 2, (255,255,255), 2, cv2.LINE_AA)
# plt.imshow(img, cmap='gray')
# cv2.imwrite("mipt_vis/{}.png".format(i) ,img)
# i += 1
# In[]: FLIR DATASET
#folders_flir = ['/home/kenny/dgx/home/datasets/ir/flir/train/', '/home/kenny/dgx/home/datasets/ir/flir/val/', '/home/kenny/dgx/home/datasets/ir/flir/video/']
folders_flir = ['/home/datasets/ir/flir/train/', '/home/datasets/ir/flir/val/', '/home/datasets/ir/flir/video/']
# In[]:
for folder_flir in folders_flir:
with open(folder_flir + 'thermal_annotations.json') as json_file:
train_data = json.load(json_file)
annotations_flir = train_data['annotations']
images_flir = train_data['images']
for i, ann in enumerate(annotations_flir):
if ann['category_id'] > 3 or ann['category_id'] == 2:
annotations_flir[i] = None
continue
elif ann['category_id'] == 3:
ann['category_id'] = 2
objects_count += 1
ann['id'] = objects_count
ann['image_id'] = ann['image_id'] + images_count
for img in images_flir:
img['file_name'] = folder_flir + img['file_name']
img['id'] = img['id'] + images_count
images_count += len(images_flir)
if 'train' in folder_flir:
annotations_flir_train = [i for i in annotations_flir if i]
images_flir_train = images_flir
elif 'val' in folder_flir:
annotations_flir_val = [i for i in annotations_flir if i]
images_flir_val = images_flir
elif 'video' in folder_flir:
annotations_flir_video = [i for i in annotations_flir if i]
images_flir_video = images_flir
# In[]:
#train_data_global = {'annotations': annotations_flir_train + annotations_flir_val + annotations_flir_video,
# 'categories': categories_mipt,
# 'images': images_flir_train + images_flir_val + images_flir_video,
# 'info': train_data['info'],
# 'licenses': train_data['licenses']}
#
#with open('train_data_flir.json', 'w') as outfile:
# json.dump(train_data_global, outfile)
# In[]: Visualize
#for ann in annotations_global:
#
# ann = annotations_global[1226]
# img = cv2.imread(images_global[ann['image_id']]['file_name'], 0)
# bbox = ann['bbox']
# cv2.rectangle(img, (int(bbox[0]), int(bbox[1])), (int(bbox[0]+bbox[2]), int(bbox[1]+bbox[3])), (255, 0, 0), 2)
# plt.imshow(img, cmap='gray')
#
# break
# In[]:
#folders_tokyo = ['/home/kenny/dgx/home/datasets/ir/tokyo/labels/fir/', '/home/kenny/dgx/home/datasets/ir/tokyo/labels/mir/', '/home/kenny/dgx/home/datasets/ir/tokyo/labels/nir/']
folders_tokyo = ['/home/datasets/ir/tokyo/labels/fir/', '/home/datasets/ir/tokyo/labels/mir/', '/home/datasets/ir/tokyo/labels/nir/']
# In[]:
#files = [f for f in glob(folders_tokyo[0] + '*.txt', recursive=True)]
#files.sort()
#for file in tqdm(files):
## file = files[4]
# img = cv2.imread(file.replace('labels', 'Images').replace('.txt', '.png'))
# with open(file) as f:
# for line in f:
# cl, x, y, w, h = line[:-1].split(" ")
# if int(cl) == 1:
# w = float(w)*640
# h = float(h)*480
# x1 = int(float(x)*640 - w/2)
# y1 = int(float(y)*480 - h/2)
# x2 = int(x1+w)
# y2 = int(y1+h)
# cv2.rectangle(img, (x1, y1), (x2, y2), (255, 0, 0), 2)
# break
#
#plt.imshow(img)
# In[]:
# 0 - person
# 1 - car
for folder_tokyo in folders_tokyo:
files = [f for f in glob(folder_tokyo + '*.txt', recursive=True)]
files.sort()
annotations_tokyo = []
images_tokyo = []
width = 640 if 'fir' in folder_tokyo else 320
height = 480 if 'fir' in folder_tokyo else 256
for ind, file in tqdm(enumerate(files)):
with open(file) as f:
for line in f:
cl, x, y, w, h = line[:-1].split(" ")
if int(cl) < 2:
objects_count += 1
w = float(w)*width
h = float(h)*height
x = int(float(x)*width - w/2)
y = int(float(y)*height - h/2)
annotation = {'area': w*h,
'bbox': [x, y, int(w), int(h)],
'category_id': int(cl) + 1,
'id': objects_count,
'image_id': ind + images_count,
'iscrowd': 0,
'segmentation': [[x, y, x, y + int(h), x + int(w), y + int(h), x + int(w), y]]
}
annotations_tokyo.append(annotation)
image = {'file_name': file.replace('labels', 'Images').replace('.txt', '.png'),
'height': height,
'id': ind + images_count,
'width': width,
}
images_tokyo.append(image)
images_count += len(images_tokyo)
if 'fir' in folder_tokyo:
annotations_tokyo_fir = annotations_tokyo
images_tokyo_fir = images_tokyo
elif 'mir' in folder_tokyo:
annotations_tokyo_mir = annotations_tokyo
images_tokyo_mir = images_tokyo
elif 'nir' in folder_tokyo:
annotations_tokyo_nir = annotations_tokyo
images_tokyo_nir = images_tokyo
# In[]:
#train_data_global = {'annotations': annotations_tokyo_fir + annotations_tokyo_mir + annotations_tokyo_nir,
# 'categories': categories_mipt,
# 'images': images_tokyo_fir + images_tokyo_mir + images_tokyo_nir,
# 'info': train_data['info'],
# 'licenses': train_data['licenses']}
#
#with open('train_data_tokyo.json', 'w') as outfile:
# json.dump(train_data_global, outfile)
# In[]:
annotations_global = annotations_mipt + annotations_flir_train + annotations_flir_val + annotations_flir_video + annotations_tokyo_fir + annotations_tokyo_mir + annotations_tokyo_nir
images_global = images_mipt + images_flir_train + images_flir_val + images_flir_video + images_tokyo_fir + images_tokyo_mir + images_tokyo_nir
categories_global = categories_mipt
del(annotations_mipt, annotations_flir, annotations_flir_train, annotations_flir_val, annotations_flir_video, annotations_tokyo_fir, annotations_tokyo_mir, annotations_tokyo_nir)
del(images_mipt, images_flir, images_flir_train, images_flir_val, images_flir_video, images_tokyo_fir, images_tokyo_mir, images_tokyo_nir)
train_data_global = {'annotations': annotations_global,
'categories': categories_global,
'images': images_global,
'info': train_data['info'],
'licenses': train_data['licenses']}
with open('train_data_3ds.json', 'w') as outfile:
json.dump(train_data_global, outfile)