forked from akshitac8/Generative_MLZSL
-
Notifications
You must be signed in to change notification settings - Fork 0
/
util.py
309 lines (264 loc) · 13.5 KB
/
util.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Fri May 8 21:56:19 2020
@author: akshitac8
"""
import torch
import torch.nn.functional as F
from sklearn import preprocessing
from sklearn.preprocessing import normalize
from sklearn.neighbors import NearestNeighbors
import os
import pickle
import h5py
import time
import numpy as np
import random
random.seed(3483)
np.random.seed(3483)
def load_dict_from_hdf5(filename):
"""
....
"""
with h5py.File(filename, 'r') as h5file:
return recursively_load_dict_contents_from_group(h5file, '/')
def recursively_load_dict_contents_from_group(h5file, path):
"""
....
"""
ans = {}
for key, item in h5file[path].items():
if isinstance(item, h5py._hl.dataset.Dataset):
ans[key] = item.value
elif isinstance(item, h5py._hl.group.Group):
ans[key] = recursively_load_dict_contents_from_group(
h5file, path + key + '/')
return ans
def mkdir(path):
if not os.path.exists(path):
os.makedirs(path)
def weights_init(m):
classname = m.__class__.__name__
if classname.find('Linear') != -1:
m.weight.data.normal_(0.0, 0.02)
if m.bias is not None:
m.bias.data.fill_(0)
elif classname.find('BatchNorm') != -1:
m.weight.data.normal_(1.0, 0.02)
m.bias.data.fill_(0)
# SYNTHESIS LABELS FROM TRAIN DATA
def generate_fake_test_from_train_labels(train_seen_label, attribute, seenclasses, unseenclasses, num, per_seen=0.10, \
per_unseen=0.40, per_seen_unseen= 0.50):
"""
Input:
train_seen_label-> images with labels containing objects less than opt.N
attribute-> array containing word embeddings
seenclasses-> array containing seen class indices
unseenclasses-> array containing unseen class indices
num-> number of generated synthetic labels
Output:
gzsl -> tensor containing synthetic labels of only unseen, seen and seen-unseen classes.
"""
if train_seen_label.min() == 0:
print("Training data already trimmed and converted")
else:
print("original training data received (-1,1)'s ")
train_seen_label = torch.clamp(train_seen_label,0,1)
#remove all zero labeled images while training
train_seen_label = train_seen_label[(train_seen_label.sum(1) != 0).nonzero().flatten()]
seen_attributes = attribute[seenclasses]
unseen_attributes = attribute[unseenclasses]
seen_percent, unseen_percent, seen_unseen_percent = per_seen , per_unseen, per_seen_unseen
print("seen={}, unseen={}, seen-unseen={}".format(seen_percent, unseen_percent, seen_unseen_percent))
print("syn num={}".format(num))
gzsl = []
for i in range(0, num):
new_gzsl_syn_list = []
seen_unseen_label_pairs = {}
nbrs = NearestNeighbors(n_neighbors=1, algorithm='auto').fit(unseen_attributes)
for seen_idx, seen_att in zip(seenclasses,seen_attributes):
_, indices = nbrs.kneighbors(seen_att[None,:])
seen_unseen_label_pairs[seen_idx.tolist()] = unseenclasses[indices[0][0]].tolist()
#ADDING ONLY SEEN LABELS
idx = torch.randperm(len(train_seen_label))[0:int(len(train_seen_label)*seen_percent)]
seen_labels = train_seen_label[idx]
_new_gzsl_syn_list = torch.zeros(seen_labels.shape[0], attribute.shape[0])
_new_gzsl_syn_list[:,:len(seenclasses)] = seen_labels
new_gzsl_syn_list.append(_new_gzsl_syn_list)
#ADDING ONLY UNSEEN LABELS
idx = torch.randperm(len(train_seen_label))[0:int(len(train_seen_label)*unseen_percent)]
temp_label = train_seen_label[idx]
_new_gzsl_syn_list = torch.zeros(temp_label.shape[0], attribute.shape[0])
for m,lab in enumerate(temp_label):
new_lab = torch.zeros(attribute.shape[0])
unseen_lab = lab.nonzero().flatten()
u=[]
for i in unseen_lab:
u.append(seen_unseen_label_pairs[i.tolist()])
new_lab[u]=1
_new_gzsl_syn_list[m,:] = new_lab
unseen_labels = _new_gzsl_syn_list
new_gzsl_syn_list.append(unseen_labels)
#ADDING BOTH SEEN AND UNSEEN LABELS 50% OF THE SELECTED SEEN LABELS IS MAPPED TO UNSEEN LABELS
idx = torch.randperm(len(train_seen_label))[0:int(len(train_seen_label)*seen_unseen_percent)]
temp_label = train_seen_label[idx]
_new_gzsl_syn_list = torch.zeros(temp_label.shape[0], attribute.shape[0])
for m,lab in enumerate(temp_label):
u = []
new_lab = torch.zeros(attribute.shape[0])
seen_unseen_lab = lab.nonzero().flatten()
temp_seen_label = np.random.choice(seen_unseen_lab,int(len(seen_unseen_lab)*0.50))
u.extend(temp_seen_label)
rem_seen_label = np.setxor1d(temp_seen_label,seen_unseen_lab)
for i in rem_seen_label:
u.append(seen_unseen_label_pairs[i.tolist()])
new_lab[u]=1
_new_gzsl_syn_list[m,:] = new_lab
seen_unseen_labels = _new_gzsl_syn_list
new_gzsl_syn_list.append(seen_unseen_labels)
new_gzsl_syn_list = torch.cat(new_gzsl_syn_list)
gzsl.append(new_gzsl_syn_list)
gzsl = torch.cat(gzsl)
tmp_list = gzsl.sum(0)
## To make sure every unseen label gets covered
empty_lab = torch.arange(tmp_list.numel())[tmp_list==0]
min_uc = int(tmp_list[len(seenclasses):][tmp_list[len(seenclasses):]>0].min().item())
for el in empty_lab:
idx = torch.randperm(gzsl.size(0))[:min_uc]
gzsl[idx,el] = 1
gzsl = gzsl.long()
print("GZSL TEST LABELS:",gzsl.shape)
return gzsl
def get_seen_unseen_classes(file_tag1k, file_tag81):
"""
Input:
file_tag1k -> NUS-WIDE provided Taglist of 1000 categories.
file_tag81 -> NUS-WIDE provided Taglist of 81 categories.
Output:
seen_cls_idx -> selected seen class indices
unseen_cls_idx -> selected unseen class indices
"""
with open(file_tag1k, "r") as file:
tag1k = np.array(file.read().splitlines())
with open(file_tag81, "r") as file:
tag81 = np.array(file.read().splitlines())
seen_cls_idx = np.array(
[i for i in range(len(tag1k)) if tag1k[i] not in tag81])
unseen_cls_idx = np.array(
[i for i in range(len(tag1k)) if tag1k[i] in tag81])
return seen_cls_idx, unseen_cls_idx
class DATA_LOADER(object):
def __init__(self, opt):
self.read_matdataset(opt)
self.index_in_epoch = 0
self.epochs_completed = 0
def read_matdataset(self, opt):
tic = time.time()
src = "datasets/NUS-WIDE" #folder for path containing features
att_path = os.path.join(src,'word_embedding','NUS_WIDE_pretrained_w2v_glove-wiki-gigaword-300')
file_tag1k = os.path.join(src,'NUS_WID_Tags','TagList1k.txt')
file_tag81 = os.path.join(src,'ConceptsList','Concepts81.txt')
self.seen_cls_idx, _ = get_seen_unseen_classes(file_tag1k, file_tag81)
src_att = pickle.load(open(att_path, 'rb'))
print("attributes are combined in this order-> seen+unseen")
self.attribute = torch.from_numpy(normalize(np.concatenate((src_att[0][self.seen_cls_idx],src_att[1]),axis=0)))
#VGG features path
import pdb;pdb.set_trace()
train_loc = load_dict_from_hdf5(os.path.join(src, 'nus_wide_vgg_features','nus_seen_train_vgg19.h5'))
test_unseen_loc = load_dict_from_hdf5(os.path.join(src, 'nus_wide_vgg_features', 'nus_zsl_test_vgg19.h5'))
test_seen_unseen_loc = load_dict_from_hdf5(os.path.join(src, 'nus_wide_vgg_features', 'nus_gzsl_test_vgg19.h5'))
feature_train_loc = train_loc['features']
label_train_loc = train_loc['labels']
feature_test_unseen_loc = test_unseen_loc['features']
label_test_unseen_loc = test_unseen_loc['labels']
feature_test_seen_unseen_loc = test_seen_unseen_loc['features']
label_test_seen_unseen_loc = test_seen_unseen_loc['labels']
print("Data loading finished, Time taken: {}".format(time.time()-tic))
tic = time.time()
if not opt.validation:
if opt.preprocessing:
if opt.standardization:
print('standardization...')
scaler = preprocessing.StandardScaler()
else:
scaler = preprocessing.MinMaxScaler()
_train_feature = scaler.fit_transform(feature_train_loc)
_test_unseen_feature = scaler.transform(feature_test_unseen_loc)
_test_seen_unseen_feature = scaler.transform(feature_test_seen_unseen_loc)
self.train_feature = torch.from_numpy(_train_feature).float()
mx = self.train_feature.max()
self.train_feature.mul_(1/mx)
self.train_label = torch.from_numpy(label_train_loc).long()
self.test_unseen_feature = torch.from_numpy(_test_unseen_feature).float()
self.test_unseen_feature.mul_(1/mx)
self.test_unseen_label = torch.from_numpy(label_test_unseen_loc).long()
self.test_seen_unseen_feature = torch.from_numpy(_test_seen_unseen_feature).float()
self.test_seen_unseen_feature.mul_(1/mx)
self.test_seen_unseen_label = torch.from_numpy(label_test_seen_unseen_loc).long()
else:
self.train_feature = torch.from_numpy(feature_train_loc).float()
self.train_label = torch.from_numpy(label_train_loc).long()
self.test_unseen_feature = torch.from_numpy(feature_test_unseen_loc).float()
self.test_unseen_label = torch.from_numpy(label_test_unseen_loc).long()
print("REMOVING ZEROS LABELS")
temp_label = torch.clamp(self.train_label,0,1)
temp_seen_labels = temp_label.sum(1)
temp_label = temp_label[temp_seen_labels>0]
self.train_label = temp_label
self.train_feature = self.train_feature[temp_seen_labels>0]
self.train_trimmed_label = self.train_label[temp_label.sum(1)<=opt.N]
self.train_trimmed_feature = self.train_feature[temp_label.sum(1)<=opt.N]
print("Data with N={} labels={}".format(opt.N,self.train_trimmed_label.shape))
print("Full Data labels={} with min label/feature = {} and max label/feature = {}".format(self.train_label.shape, temp_label.sum(1).min(), temp_label.sum(1).max()))
self.seenclasses = torch.from_numpy(np.arange(0, self.seen_cls_idx.shape[-1])) # [0-925]
self.unseenclasses = torch.from_numpy(np.arange(0+self.seen_cls_idx.shape[-1], len(self.attribute))) # [925-1006]
self.N = opt.N
self.syn_num = opt.syn_num
self.per_seen = opt.per_seen
self.per_unseen = opt.per_unseen
self.per_seen_unseen = opt.per_seen_unseen
print("USING TRAIN FEATURES WITH <=N")
self.ntrain = self.train_trimmed_feature.size()[0]
train_labels = self.train_trimmed_label
self.ntest_unseen = self.test_unseen_feature.size()[0]
self.ntrain_class = self.seenclasses.size(0)
self.ntest_class = self.unseenclasses.size(0)
self.train_class = self.seenclasses.clone()
self.allclasses = torch.arange(0, self.ntrain_class + self.ntest_class).long()
self.GZSL_fake_test_labels = generate_fake_test_from_train_labels(train_labels, self.attribute, self.seenclasses, \
self.unseenclasses, self.syn_num, self.per_seen, self.per_unseen, self.per_seen_unseen)
print("Data preprocssing finished, Time taken: {}".format(time.time()-tic))
def _average(self, lab, attribute):
return torch.mean(attribute[lab], 0)
def ALF_preprocess_att(self, labels, attribute):
new_seen_attribute = torch.zeros(labels.shape[0], attribute.shape[-1])
for i in range(len(labels)):
lab = labels[i].nonzero().flatten()
if len(lab) == 0: continue
new_seen_attribute[i, :] = self._average(lab, attribute)
return new_seen_attribute
def FLF_preprocess_att(self, labels, attribute):
new_attributes = torch.zeros(labels.shape[0], self.N, attribute.shape[-1]) #new attributes [BS X 10 X 925]
for i in range(len(labels)):
lab = labels[i].nonzero().flatten()
if len(lab) == self.N: new_attributes[i,:,:] = attribute[lab]
elif len(lab) < self.N: new_attributes[i,:,:] = torch.cat((attribute[lab],torch.zeros((self.N - len(lab)), attribute.shape[-1])))
return new_attributes
## Training Dataloader
def next_train_batch(self, batch_size):
idx = torch.randperm(self.ntrain)[0:batch_size]
feature = self.train_trimmed_feature
labels = self.train_trimmed_label
batch_feature = feature[idx]
batch_labels = labels[idx]
early_fusion_train_batch_att = self.ALF_preprocess_att(batch_labels, self.attribute)
late_fusion_train_batch_att = self.FLF_preprocess_att(batch_labels, self.attribute)
return batch_labels, batch_feature, late_fusion_train_batch_att, early_fusion_train_batch_att
## Testing Dataloader
def next_test_batch(self, batch_size):
idx = torch.randperm(len(self.GZSL_fake_test_labels))[0:batch_size]
batch_labels = self.GZSL_fake_test_labels[idx]
early_fusion_test_batch_att = self.ALF_preprocess_att(batch_labels, self.attribute)
late_fusion_test_batch_att = self.FLF_preprocess_att(batch_labels, self.attribute)
return batch_labels, late_fusion_test_batch_att, early_fusion_test_batch_att