-
Notifications
You must be signed in to change notification settings - Fork 10
/
Copy pathtrain.py
131 lines (110 loc) · 4.54 KB
/
train.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
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.autograd import Variable
from numpy.random import normal
import numpy as np
import os
from torch.nn import init
from torch.utils import data
import torchvision
import random
from cStringIO import StringIO
import time
import scipy.misc
# from PIL import Image
from matplotlib import pyplot as plt
from dataloader import ImageNet_Dataloader
from PIL import Image
from network import netG, netD
import torchvision.utils as vutils
def weights_init_xavier(m):
classname = m.__class__.__name__
# print(classname)
if classname.find('Conv') != -1:
init.xavier_normal(m.weight.data, gain=1)
elif classname.find('Linear') != -1:
init.xavier_normal(m.weight.data, gain=1)
elif classname.find('BatchNorm2d') != -1:
init.uniform(m.weight.data, 1.0, 0.02)
init.constant(m.bias.data, 0.0)
# batch_size = 40
netD = netD().cuda(0)
netD.apply(weights_init_xavier)
netD.train(True)
netG = netG().cuda(0)
netG.apply(weights_init_xavier)
netG.train(True)
# optimizer = optim.Adam(netD.parameters(),lr=0.005, betas=(0.5, 0.999))
optimizerD = optim.Adam(netD.parameters(), lr=0.0005, betas=(0.5, 0.999))
optimizerG = optim.Adam(netG.parameters(), lr=0.0005, betas=(0.5, 0.999))
criterion = nn.BCELoss(size_average=True).cuda(0)
data_path = '/media/tensor-server/ee577d95-535d-40b2-88db-5546defabb74/imagenet20_rgbmsk_v0/ImageNet20/'
dst = ImageNet_Dataloader(data_path, is_transform=True)
print('length of the dataset', len(dst))
trainloader = data.DataLoader(dst, batch_size=24,shuffle=True)
step_index = 0
real_label = 1
fake_label = 0
for epo_num in range(500):
for i, data in enumerate(trainloader):
real_img, real_mask = data
print(real_img)
step_index = step_index + 1
real_img, real_mask = Variable(real_img).cuda(0), Variable(real_mask).cuda(0)
batch_size = real_img.size(0)
input = torch.FloatTensor(batch_size, 3, 128, 128)
noise = torch.FloatTensor(batch_size, 100, 1, 1)
fixed_noise = torch.FloatTensor(batch_size, 100, 1, 1).normal_(0, 1)
label = torch.FloatTensor(batch_size,1,16,16)
fixed_noise = Variable(fixed_noise).cuda(0)
netD.zero_grad()
# train with real data
labelv = Variable(label.fill_(real_label)).cuda(0)
fake_prob, forg_prob = netD(real_img)
real_forg_prob = forg_prob.squeeze(1)
real_forg_prob = real_forg_prob.view(-1)
real_mask = real_mask.view(-1)
seg_loss = criterion(real_forg_prob, real_mask)
errD_real = criterion(fake_prob, labelv)
D_x = fake_prob.data.mean()
# train with fake
# random initialize the noise vector
noisev = Variable(noise.normal_(0, 1)).cuda(0)
fake_img = netG(noisev)
labelv = Variable(label.fill_(fake_label)).cuda(0)
fake_prob, _ = netD(fake_img.detach())
errD_fake = criterion(fake_prob, labelv)
D_G_z1 = fake_prob.data.mean()
bce_loss = (errD_real + errD_fake)*0.5
errD = bce_loss*0.1 + seg_loss
errD.backward()
optimizerD.step()
############################
# (2) Update G network: maximize log(D(G(z)))
###########################
netG.zero_grad()
labelv = Variable(label.fill_(real_label)).cuda(0) # fake labels are real for generator cost
fake_prob, fake_forg_prob = netD(fake_img)
errG = criterion(fake_prob, labelv)
errG.backward()
D_G_z2 = fake_prob.data.mean()
optimizerG.step()
print('[%d/%d][%d/%d] Loss_D: %.4f BCE_D: %.4f SEG_D: %.4f Loss_G: %.4f D(x): %.4f D(G(z)): %.4f / %.4f'
% (epo_num, 100, step_index%len(trainloader), len(trainloader),
errD.data[0],bce_loss, seg_loss, errG.data[0], D_x, D_G_z1, D_G_z2))
if step_index % 200 == 0:
vutils.save_image(real_img.data,
'./samples/real_samples_epoch_%06d.png' % (step_index),
normalize=True)
vutils.save_image(forg_prob.data,
'./samples/real_foreground_samples_epoch_%06d.png' % (step_index),
normalize=True)
vutils.save_image(fake_forg_prob.data,
'./samples/fake_forg_prob_samples_epoch_%06d.png' % (step_index),
normalize=True)
fake = netG(fixed_noise)
vutils.save_image(fake.data,
'./samples/fake_samples_epoch_%06d.png' % (step_index),
normalize=True)