-
Notifications
You must be signed in to change notification settings - Fork 2
/
avse_main.py
418 lines (372 loc) · 17.3 KB
/
avse_main.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
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
from __future__ import print_function
import argparse
import os
import random
import torch
import torch.nn as nn
import torch.autograd as autograd
import torch.optim as optim
import torch.backends.cudnn as cudnn
from torch.autograd import Variable
import torch.nn.functional as func
import math
import util
import sys
import classifier
import classifier2
import model
from sklearn import preprocessing
import scipy.io as sio
import visdom
parser = argparse.ArgumentParser()
parser.add_argument('--dataset', default='CUB', help='dataset')
parser.add_argument('--dataroot', default='./data/', help='path to dataset')
parser.add_argument('--matdataset', default=True, help='Data in matlab format')
parser.add_argument('--image_embedding', default='res101')
parser.add_argument('--class_embedding', default='att')
parser.add_argument('--syn_num', type=int, default=300, help='number features to generate per class')
parser.add_argument('--gzsl', default=False, help='enable generalized zero-shot learning')
parser.add_argument('--preprocessing', default=True, help='enbale MinMaxScaler on visual features')
parser.add_argument('--standardization', default=False)
parser.add_argument('--validation', default=False, help='enable cross validation mode')
parser.add_argument('--workers', type=int, help='number of data loading workers', default=2)
parser.add_argument('--batch_size', type=int, default=64, help='input batch size')
parser.add_argument('--resSize', type=int, default=2048, help='size of visual features')
parser.add_argument('--attSize', type=int, default=312, help='size of semantic features')
parser.add_argument('--nz', type=int, default=312, help='size of the latent z vector')
parser.add_argument('--ngh', type=int, default=4096, help='size of the hidden units in generator')
parser.add_argument('--ndh', type=int, default=4096, help='size of the hidden units in discriminator')
parser.add_argument('--nepoch', type=int, default=100, help='number of epochs to train for')
parser.add_argument('--critic_iter', type=int, default=5, help='critic iteration, following WGAN-GP')
parser.add_argument('--lambda1', type=float, default=10, help='gradient penalty regularizer, following WGAN-GP')
parser.add_argument('--cls_weight', type=float, default=0.01, help='weight of the classification loss')
parser.add_argument('--gamma', type=float, default=0.1, help='weight of the regression loss')
parser.add_argument('--lr', type=float, default=0.0001, help='learning rate to train GANs ')
parser.add_argument('--classifier_lr', type=float, default=0.001, help='learning rate to train softmax classifier')
parser.add_argument('--beta1', type=float, default=0.5, help='beta1 for adam. default=0.5')
parser.add_argument('--cuda', default=True, help='enables cuda')
parser.add_argument('--ngpu', type=int, default=1, help='number of GPUs to use')
parser.add_argument('--pretrain_classifier', default='', help="path to pretrain classifier (to continue training)")
parser.add_argument('--netG', default='', help="path to netG (to continue training)")
parser.add_argument('--netD', default='', help="path to netD (to continue training)")
parser.add_argument('--netG_name', default='MLP_G')
parser.add_argument('--netD_name', default='MLP_CRITIC')
parser.add_argument('--save_path', default='./checkpoint/', help='folder to output data and model checkpoints')
parser.add_argument('--save_name', default='CUB', help='folder to output data and model checkpoints')
parser.add_argument('--save_every', type=int, default=50)
parser.add_argument('--print_every', type=int, default=1)
parser.add_argument('--val_every', type=int, default=10)
parser.add_argument('--start_epoch', type=int, default=0)
parser.add_argument('--manualSeed', type=int, default=3483, help='manual seed')
parser.add_argument('--nclass_all', type=int, default=200, help='number of all classes')
#
#parser.add_argument('--fc1_size', type=int, default=4096, help='the first FC layer')
#parser.add_argument('--fc2_size', type=int, default=512, help='the second FC layer')
opt = parser.parse_args()
print(opt)
try:
os.makedirs(opt.save_path)
except OSError:
pass
if opt.manualSeed is None:
opt.manualSeed = random.randint(1, 10000)
print("Random Seed: ", opt.manualSeed)
random.seed(opt.manualSeed)
torch.manual_seed(opt.manualSeed)
if opt.cuda:
torch.cuda.manual_seed_all(opt.manualSeed)
# cudnn.benchmark may help increase the running time.
cudnn.benchmark = True
ngpu = 1
# Decide which device we want to run on
device = torch.device("cuda" if (torch.cuda.is_available() and ngpu > 0) else "cpu")
if torch.cuda.is_available() and not opt.cuda:
print("WARNING: You have a CUDA device, so you should probably run with --cuda")
# load data, this is a class instance
data = util.DATA_LOADER(opt)
print("# of training samples: ", data.ntrain)
# image encoder
netG_Img = model.MLP_G_Img_Adaptive(opt)
netG_Img.apply(model.weights_init)
print(netG_Img)
# generator
netG_Att = model.MLP_G_Att_Adaptive(opt)
netG_Att.apply(model.weights_init)
print(netG_Att)
# regressor (decoder)
netE = model.MLP_E_Adaptive(opt)
netE.apply(model.weights_init)
print(netE)
# discriminator
netD = model.MLP_D(opt)
netD.apply(model.weights_init)
print(netD)
# classification loss, Equation (4) of the paper
cls_criterion = nn.NLLLoss()
mse_criterion = nn.MSELoss()
l1_criterion = nn.L1Loss()
sim_criterion = nn.CosineEmbeddingLoss()
input_res = torch.FloatTensor(opt.batch_size, opt.resSize)
input_att = torch.FloatTensor(opt.batch_size, opt.attSize)
noise = torch.FloatTensor(opt.batch_size, opt.nz)
one = torch.FloatTensor([1])
mone = one * -1
input_label = torch.LongTensor(opt.batch_size)
noise_cls = torch.FloatTensor(opt.nclass_all, opt.nz)
noise_unseen = torch.FloatTensor(data.ntest_class, opt.nz)
unseen_att = torch.FloatTensor(data.ntest_class, opt.attSize)
unseen_label = torch.LongTensor(data.ntest_class)
one_batch = torch.ones(opt.batch_size).float()
one_batch_unseen = torch.ones(data.ntest_class).long()
zero_batch = torch.zeros(opt.batch_size).long()
fixed_noise = torch.FloatTensor(opt.syn_num, opt.nz).normal_(0, 1)
if opt.cuda:
netD.cuda()
netG_Img.cuda()
netG_Att.cuda()
netE.cuda()
input_res = input_res.cuda()
noise, input_att = noise.cuda(), input_att.cuda()
one = one.cuda()
mone = mone.cuda()
cls_criterion.cuda()
mse_criterion.cuda()
sim_criterion.cuda()
l1_criterion.cuda()
input_label = input_label.cuda()
noise_cls = noise_cls.cuda()
noise_unseen = noise_unseen.cuda()
unseen_att = unseen_att.cuda()
unseen_label = unseen_label.cuda()
one_batch = one_batch.cuda()
zero_batch = zero_batch.cuda()
one_batch_unseen = one_batch_unseen.cuda()
fixed_noise = fixed_noise.cuda()
if opt.standardization:
print('standardization...')
scaler = preprocessing.StandardScaler()
else:
scaler = preprocessing.MinMaxScaler()
def sample():
batch_feature, batch_label, batch_att = data.next_batch(opt.batch_size)
input_res.copy_(batch_feature)
input_att.copy_(batch_att)
input_label.copy_(util.map_label(batch_label, data.seenclasses))
def generate_syn_feature(netG, classes, attribute, num):
nclass = classes.size(0)
syn_feature = torch.FloatTensor(nclass * num, opt.resSize)
syn_label = torch.LongTensor(nclass * num)
syn_att = torch.FloatTensor(num, opt.attSize)
#syn_noise = torch.FloatTensor(num, opt.nz)
if opt.cuda:
syn_att = syn_att.cuda()
#syn_noise = syn_noise.cuda()
for i in range(nclass):
iclass = classes[i]
iclass_att = attribute[iclass]
syn_att.copy_(iclass_att.repeat(num, 1))
#syn_noise.normal_(0, 1)
with torch.no_grad():
output = netG(fixed_noise, syn_att)
syn_feature.narrow(0, i * num, num).copy_(output.data.cpu())
syn_label.narrow(0, i * num, num).fill_(iclass)
return syn_feature, syn_label
def compute_sim_loss(img_embed, text_embed):
sim_loss = sim_criterion(img_embed, text_embed, one_batch)
return sim_loss
def calc_gradient_penalty(netD, real_data, fake_data, input_att):
# print real_data.size()
alpha = torch.rand(opt.batch_size, 1)
alpha = alpha.expand(real_data.size())
if opt.cuda:
alpha = alpha.cuda()
interpolates = alpha * real_data + ((1 - alpha) * fake_data)
if opt.cuda:
interpolates = interpolates.cuda()
interpolates = Variable(interpolates, requires_grad=True)
disc_interpolates = netD(interpolates, Variable(input_att))
ones = torch.ones(disc_interpolates.size())
if opt.cuda:
ones = ones.cuda()
gradients = autograd.grad(outputs=disc_interpolates, inputs=interpolates,
grad_outputs=ones,
create_graph=True, retain_graph=True, only_inputs=True)[0]
gradient_penalty = ((gradients.norm(2, dim=1) - 1) ** 2).mean() * opt.lambda1
return gradient_penalty
def KL_loss(mu, logvar):
# -0.5 * sum(1 + log(sigma^2) - mu^2 - sigma^2)
KLD_element = mu.pow(2).add_(logvar.exp()).mul_(-1).add_(1).add_(logvar)
KLD = torch.mean(KLD_element).mul_(-0.5)
return KLD
print("Starting to train classifier...")
# train a classifier on seen classes, obtain \theta of Equation (4)
pretrain_cls = classifier.CLASSIFIER(data, data.train_feature, util.map_label(data.train_label, data.seenclasses),
data.seenclasses.size(0), opt.resSize, opt.cuda, opt.classifier_lr, 0.5, 100, 100,
opt.pretrain_classifier)
# for p in pretrain_cls.model.parameters(): # set requires_grad to False
# p.requires_grad = False
print("Starting to train loop...")
# setup optimizer
optimizerD = optim.Adam(netD.parameters(), lr=opt.lr, betas=(opt.beta1, 0.999))
optimizerG = optim.Adam(list(netG_Img.parameters())+list(netG_Att.parameters())+
list(pretrain_cls.model.parameters())+list(netE.parameters()), lr=opt.lr, betas=(opt.beta1, 0.999))
#optimizerU = optim.Adam(list(netG_Att.parameters())+list(netC.parameters()), lr=opt.lr, betas=(opt.beta1, 0.999))
# visdom setup
#vis = util.Visualizer(env='LatGAN')
###
best_acc_zsl = 0.0
best_acc_unseen_gzsl = 0.0
best_acc_seen_gzsl = 0.0
best_H_gzsl = 0.0
opt.n_batch = data.ntrain // opt.batch_size
# freeze the classifier during the optimization
for epoch in range(opt.nepoch):
## random shuffle
idx = torch.randperm(data.ntrain)
data.train_feature = data.train_feature[idx]
data.train_label = data.train_label[idx]
##
for i in range(opt.n_batch):
start_i = i * opt.batch_size
end_i = start_i + opt.batch_size
batch_feature = data.train_feature[start_i:end_i]
batch_label = data.train_label[start_i:end_i]
batch_att = data.attribute[batch_label]
# copy data to GPU
input_res.copy_(batch_feature)
input_att.copy_(batch_att)
input_label.copy_(util.map_label(batch_label, data.seenclasses))
############################
# (1) Update D network
###########################
# for p in netD.parameters(): # reset requires_grad
# p.requires_grad = True # they are set to False below in netG update
for iter_d in range(opt.critic_iter):
## call a function!!!
# sample()
netD.zero_grad()
# train with realG
real_res = netG_Img(input_res)
criticD_real = netD(real_res, input_att).view(-1)
criticD_real = criticD_real.mean()
# criticD_real.backward(mone)
# train with fakeG
noise.normal_(0, 1)
fake_res = netG_Att(noise, input_att)
criticD_fake = netD(fake_res, input_att) ## detach???
criticD_fake = criticD_fake.mean()
# gradient penalty
gradient_penalty = calc_gradient_penalty(netD, real_res.data, fake_res.data, input_att)
##
D_gan_loss = criticD_fake - criticD_real
D_loss = D_gan_loss + gradient_penalty
D_loss.backward()
optimizerD.step()
############################
# (2) Update G network
###########################
netG_Img.zero_grad()
netG_Att.zero_grad()
netE.zero_grad()
pretrain_cls.model.zero_grad()
###
real_res = netG_Img(input_res)
noise.normal_(0, 1)
fake_res = netG_Att(noise, input_att)
criticG_fake = netD(fake_res, input_att)
criticG_fake = criticG_fake.mean()
G_gan_loss = -criticG_fake ### ??? remove real?
# classification loss
G_cls_loss_att = cls_criterion(pretrain_cls.model(fake_res), input_label)
G_cls_loss_img = cls_criterion(pretrain_cls.model(real_res), input_label)
G_cls_loss = G_cls_loss_att + G_cls_loss_img
# reconst loss
#real_res = netG_Img(input_res)
real_rect = netE(real_res)
G_rect_loss_img = l1_criterion(real_rect, input_att)
fake_rect = netE(fake_res)
G_rect_loss_att = l1_criterion(fake_rect, input_att)
# total loss
G_loss = G_gan_loss + opt.cls_weight * G_cls_loss + opt.gamma * G_rect_loss_att + G_rect_loss_img
G_loss.backward()
optimizerG.step()
# vis.plot_loss({'D_gan_loss': D_gan_loss.item(),
# 'G_gan_loss': G_gan_loss.item(), 'G_cls_loss_att': G_cls_loss_att.item(), 'G_cls_loss_img': G_cls_loss_img.item(),
# 'G_rect_loss_att': G_rect_loss_att.item()},
# title='LatGAN', xlabel='Epochs', ylabel='Training loss')
#evaluate the model, set G to evaluation mode
netG_Img.eval()
netG_Att.eval()
#Generalized zero-shot learning
if opt.gzsl:
## test unseen img embed
ntest = data.ntest_unseen
nclass = data.ntest_class
test_unseen_img_embed = torch.FloatTensor(ntest, opt.resSize)
# test_text_embed = torch.FloatTensor(nclass, opt.embedSize)
start = 0
for ii in range(0, ntest, opt.batch_size):
end = min(ntest, start + opt.batch_size)
test_feature = data.test_unseen_feature[start:end]
if opt.cuda:
test_feature = test_feature.cuda()
img_embed = netG_Img(test_feature)
test_unseen_img_embed[start:end, :] = img_embed.data.cpu()
start = end
## synthesize unseen image embed
syn_feature, syn_label = generate_syn_feature(netG_Att, data.unseenclasses, data.attribute, opt.syn_num)
train_X = torch.cat((data.train_feature, syn_feature), 0)
train_Y = torch.cat((data.train_label, syn_label), 0)
train_X = torch.cat((train_X, syn_feature), 0)
train_Y = torch.cat((train_Y, syn_label), 0)
nclass = opt.nclass_all
cls = classifier2.CLASSIFIER(train_X, train_Y, data, nclass, opt.cuda, opt.classifier_lr, 0.5, 50, opt.syn_num,
True, test_unseen_feature=test_unseen_img_embed, test_seen_feature=data.test_seen_feature)
print('%.4f %.4f %.4f' % (cls.acc_unseen, cls.acc_seen, cls.H))
if cls.H > best_H_gzsl:
best_H_gzsl = cls.H
best_acc_unseen_gzsl = cls.acc_unseen
best_acc_seen_gzsl = cls.acc_seen
# Zero-shot learning
else:
syn_feature, syn_label = generate_syn_feature(netG_Att, data.unseenclasses, data.attribute, opt.syn_num)
## extract visual latent features
# extract image embed for unseen classes
ntest = data.ntest_unseen
nclass = data.ntest_class
test_img_embed = torch.FloatTensor(ntest, opt.resSize)
# test_text_embed = torch.FloatTensor(nclass, opt.embedSize)
start = 0
for ii in range(0, ntest, opt.batch_size):
end = min(ntest, start + opt.batch_size)
test_feature = data.test_unseen_feature[start:end]
if opt.cuda:
test_feature = test_feature.cuda()
img_embed = netG_Img(test_feature)
test_img_embed[start:end, :] = img_embed.data.cpu()
start = end
cls = classifier2.CLASSIFIER(syn_feature, util.map_label(syn_label, data.unseenclasses), data,
data.unseenclasses.size(0), opt.cuda, opt.classifier_lr, 0.5, 50, opt.syn_num,
False, test_unseen_feature=test_img_embed, test_seen_feature=None) # opt.syn_num
acc = cls.acc
#print('unseen class accuracy= ', acc)
print('[%d/%d] unseen class accuracy = %.4f' % (epoch, opt.nepoch, acc))
#vis.plot_acc({'test unseen': acc}, title='debug', xlabel='Epochs', ylabel='Top-1 accuracy')
if acc > best_acc_zsl:
best_acc_zsl = acc
print('Save model!')
torch.save(netG_Img.state_dict(),
os.path.join(opt.save_path, opt.save_name + '_AVSE:netG_Img_epoch_%d.ckpt'%(epoch)))
torch.save(netG_Att.state_dict(),
os.path.join(opt.save_path, opt.save_name + '_AVSE:netG_Att_epoch_%d.ckpt'%(epoch)))
torch.save(netD.state_dict(),
os.path.join(opt.save_path, opt.save_name + '_AVSE:netD_epoch_%d.ckpt'%(epoch)))
torch.save(netE.state_dict(),
os.path.join(opt.save_path, opt.save_name + '_AVSE:netE_epoch_%d.ckpt'%(epoch)))
# reset G to training mode
netG_Img.train()
netG_Att.train()
if opt.gzsl:
print('GZSL best unseen=%.4f, seen=%.4f, h=%.4f' % (best_acc_unseen_gzsl, best_acc_seen_gzsl, best_H_gzsl))
else:
print('ZSL best unseen=%.4f' % (best_acc_zsl))