-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmake_cub_data.py
312 lines (223 loc) · 10.7 KB
/
make_cub_data.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
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
import argparse
import os
import random
import numpy as np
import pandas as pd
def _check_dir(path):
return os.path.isdir(path)
def make_dataset(dir_path, split=0.0, label=0, second_split=-1.0):
dirs = os.listdir(dir_path)
dir_path = dir_path[14:]
to_add = len(dirs)
paths = []
senond_split = []
if second_split < 0:
for dir in dirs:
if random.random() >= split:
paths.append((os.path.join(dir_path, dir), label))
else:
senond_split.append((os.path.join(dir_path, dir), label))
return to_add, paths, senond_split
else:
third_split = []
for dir in dirs:
rand = random.random()
if rand >= (split + second_split):
paths.append((os.path.join(dir_path, dir), label)) # train
elif split <= rand < (second_split + split):
senond_split.append((os.path.join(dir_path, dir), label)) # val
else:
third_split.append((os.path.join(dir_path, dir), label)) # test
return to_add, paths, senond_split, third_split
def save_dataset(ls, data_path, name):
df = pd.DataFrame({'image': list(list(zip(*ls))[0]), 'label': list(list(zip(*ls))[1])})
df.to_csv(os.path.join(data_path, f'cub_{name}.csv'), index=False, header=True)
def load_cub_data(args):
split_path = args.split_path
data_path = args.data_path
version = args.version
with open(os.path.join(split_path, f'testclasses{version}.txt')) as f:
test_cls = f.read().split('\n')[:-1]
with open(os.path.join(split_path, f'trainclasses{version}.txt')) as f:
train_cls = f.read().split('\n')[:-1]
with open(os.path.join(split_path, f'valclasses{version}.txt')) as f:
val_cls = f.read().split('\n')[:-1]
tr = []
kwn_vl = []
ukwn_vl = []
kwn_ts = []
ukwn_ts = []
total_train = 0
total_val = 0
total_test = 0
for train in train_cls:
lbl = int(train[:3])
ln, tr_r, kwn_vl_r, kwn_ts_r = make_dataset(os.path.join(data_path, train),
label=lbl,
split=0.33, # test
second_split=0.12) # val
total_train += ln
tr.extend(tr_r)
kwn_vl.extend(kwn_vl_r)
kwn_ts.extend(kwn_ts_r)
for val in val_cls:
lbl = int(val[:3])
ln, ukwn_vl_r, kwn_ts_r = make_dataset(os.path.join(data_path, val), label=lbl, split=0.1)
total_val += ln
ukwn_vl.extend(ukwn_vl_r)
kwn_ts.extend(kwn_ts_r)
for test in test_cls:
lbl = int(test[:3])
ln, ukwn_ts_r, temp = make_dataset(os.path.join(data_path, test), label=lbl, split=0)
assert len(temp) == 0
total_test += ln
ukwn_ts.extend(ukwn_ts_r)
output_string = ''
print("real total test: ", total_test)
output_string += "real total test: " + str(total_test) + '\n'
print("unknown test: ", len(ukwn_ts))
output_string += "unknown test: " + str(len(ukwn_ts)) + '\n'
print("known test:", len(kwn_ts))
output_string += "known test: " + str(len(kwn_ts)) + '\n'
print("Total test set:", len(ukwn_ts) + len(kwn_ts))
output_string += "Total test set: " + str((len(ukwn_ts) + len(kwn_ts))) + '\n'
print('*' * 30)
output_string += ('*' * 30) + '\n'
print("real total val: ", total_val)
output_string += "real total val: " + str(total_val) + '\n'
print("unknown val: ", len(ukwn_vl))
output_string += "unknown val: " + str(len(ukwn_vl)) + '\n'
print("known val:", len(kwn_vl))
output_string += "known val: " + str(len(kwn_vl)) + '\n'
print("Total val set:", len(kwn_vl) + len(ukwn_vl))
output_string += "Total val set: " + str(len(kwn_vl) + len(ukwn_vl)) + '\n'
print('*' * 30)
output_string += ('*' * 30) + '\n'
print("real train val: ", total_train)
output_string += "real train val: " + str(total_train) + '\n'
print("Total train set:", len(tr))
output_string += "Total train set: " + str(len(tr)) + '\n'
print('*' * 30)
output_string += ('*' * 30) + '\n'
print("Total trainval set:", len(tr) + len(kwn_vl) + len(ukwn_vl))
output_string += "Total trainval set: " + str(len(tr) + len(kwn_vl) + len(ukwn_vl)) + '\n'
input()
if not os.path.exists(os.path.join(data_path, f'newsplits{version}_{args.save_version}')):
data_path = os.path.join(data_path, f'newsplits{version}_{args.save_version}')
os.mkdir(data_path)
else:
data_path = os.path.join(data_path, f'newsplits{version}_{args.save_version}')
with open(os.path.join(data_path, 'config.txt'), 'w') as f:
f.write(output_string)
save_dataset(tr, data_path, 'train')
save_dataset(kwn_vl, data_path, 'knwn_cls_val')
save_dataset(ukwn_vl, data_path, 'uknwn_cls_val')
save_dataset(kwn_ts, data_path, 'knwn_cls_test')
save_dataset(ukwn_ts, data_path, 'uknwn_cls_test')
_number_of_classes(data_path)
def _number_of_classes(datapath):
print('data split path', datapath)
dirs = os.listdir(datapath)
csvs = []
names = []
for dir in dirs:
if dir.endswith('.csv'):
df = pd.read_csv(os.path.join(datapath, dir))
csvs.append(df)
names.append(dir[:-4])
# cub_train = pd.read_csv(os.path.join(datapath, 'cub_train' + '.csv'))
# cub_knwn_cls_test = pd.read_csv(os.path.join(datapath, 'cub_knwn_cls_test' + '.csv'))
# cub_knwn_cls_val = pd.read_csv(os.path.join(datapath, 'cub_knwn_cls_val' + '.csv'))
# cub_uknwn_cls_test = pd.read_csv(os.path.join(datapath, 'cub_uknwn_cls_test' + '.csv'))
# cub_uknwn_cls_val = pd.read_csv(os.path.join(datapath, 'cub_uknwn_cls_val' + '.csv'))
for name, csv in zip(names, csvs):
print(name, "number of classes:", len(np.unique(list(csv.label))))
def make_from_new_split(args):
data_path = args.data_path
splits = pd.read_csv(os.path.join(args.split_path, 'correct_split.csv'))
trainval_lbls = np.array(splits.label[splits.split == 'trainval'])
test_seen_lbls = np.array(splits.label[splits.split == 'test_seen'])
test_unseen_lbls = np.array(splits.label[splits.split == 'test_unseen'])
trainval_pths = np.array(splits.path[splits.split == 'trainval'])
test_seen_pths = np.array(splits.path[splits.split == 'test_seen'])
test_unseen_pths = np.array(splits.path[splits.split == 'test_unseen'])
test_seen = list(zip(test_seen_pths, test_seen_lbls))
test_unseen = list(zip(test_unseen_pths, test_unseen_lbls))
trainval = list(zip(trainval_pths, trainval_lbls))
assert len(np.unique(trainval_lbls)) == 150
assert len(np.unique(test_seen_lbls)) == 150
assert len(np.unique(test_unseen_lbls)) == 50
val_unseen_lbls = np.random.choice(np.unique(trainval_lbls), args.val_unseen, replace=False)
val_unseen = []
rest = []
for (path, lbl) in trainval:
if lbl in val_unseen_lbls:
val_unseen.append((path, lbl))
else:
rest.append((path, lbl))
assert len(np.unique(list(list(zip(*rest))[1]))) == 150 - (args.val_unseen)
assert len(np.unique(list(list(zip(*val_unseen))[1]))) == (args.val_unseen)
val_seen_idx = np.random.choice([i for i in range(len(rest))], int(np.floor(len(rest) * args.trainval_portion)), replace=False)
val_seen_paths = np.array(list(zip(*rest))[0])[val_seen_idx]
val_seen_lbls = np.array(list(zip(*rest))[1])[val_seen_idx]
val_seen = list(zip(val_seen_paths, val_seen_lbls))
print(val_seen[0])
train = []
for pair in rest:
if pair not in val_seen:
train.append(pair)
print('number of labels in val_seen:', len(np.unique(list(list(zip(*val_seen))[1]))))
assert len(np.unique(list(list(zip(*val_seen))[1]))) == 150 - (args.val_unseen)
assert len(np.unique(list(list(zip(*train))[1]))) == 150 - (args.val_unseen)
output_string = ''
print("unseen test:", len(test_unseen))
output_string += "known test: " + str(len(test_unseen)) + '\n'
print("seen test: ", len(test_seen))
output_string += "unknown test: " + str(len(test_seen)) + '\n'
print("Total test set:", len(test_seen) + len(test_unseen))
output_string += "Total test set: " + str((len(test_seen) + len(test_unseen))) + '\n'
print('*' * 30)
output_string += ('*' * 30) + '\n'
print("unseen val: ", len(val_unseen))
output_string += "unknown val: " + str(len(val_unseen)) + '\n'
print("seen val:", len(val_seen))
output_string += "known val: " + str(len(val_seen)) + '\n'
print("Total val set:", len(val_unseen) + len(val_seen))
output_string += "Total val set: " + str(len(val_unseen) + len(val_seen)) + '\n'
print('*' * 30)
output_string += ('*' * 30) + '\n'
print("Train set:", len(train))
output_string += "Total train set: " + str(len(train)) + '\n'
print('*' * 30)
output_string += ('*' * 30) + '\n'
print("Total trainval set:", len(train) + len(val_seen) + len(val_unseen))
output_string += "Total trainval set: " + str(len(train) + len(val_seen) + len(val_unseen)) + '\n'
input()
if not os.path.exists(os.path.join(data_path, f'final_newsplits{args.version}_{args.save_version}')):
data_path = os.path.join(data_path, f'final_newsplits{args.version}_{args.save_version}')
os.mkdir(data_path)
else:
data_path = os.path.join(data_path, f'final_newsplits{args.version}_{args.save_version}')
with open(os.path.join(data_path, 'config.txt'), 'w') as f:
f.write(output_string)
save_dataset(train, data_path, 'train')
save_dataset(val_seen, data_path, 'val_seen')
save_dataset(val_unseen, data_path, 'val_unseen')
save_dataset(test_seen, data_path, 'test_seen')
save_dataset(test_unseen, data_path, 'test_unseen')
_number_of_classes(data_path)
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-sp', '--split_path', default='CUB_splits/',help="split path")
parser.add_argument('-dp', '--data_path', default='../../dataset/CUB/images/', help="data path")
parser.add_argument('-v', '--version', type=int, default=0, help="version")
parser.add_argument('-sv', '--save_version', type=int, default=1, help="save_version")
parser.add_argument('-tvp', '--trainval_portion', type=float, default=0.2, help="percentage of seen val to trainval")
parser.add_argument('-vu', '--val_unseen', type=int, default=25, help="number of unseen val classes")
args = parser.parse_args()
# load_cub_data(args)
# make_from_new_split(args)
_number_of_classes('../../dataset/CUB/images/final_newsplits0_1')
# https://jakevdp.github.io/PythonDataScienceHandbook/04.06-customizing-legends.html
if __name__ == '__main__':
main()