-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathtrain.py
118 lines (100 loc) · 4 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
import os
import matplotlib.pyplot as plt
import numpy as np
import torch
import torchvision
import torchvision.transforms as transforms
from model import CycleGAN
import time
from arguments import Arguments
from dataset import FaceDataSet
from PIL import Image
import random
def tensor2img(img):
if not isinstance(img, np.ndarray):
if isinstance(img, torch.Tensor):
img = img.data
img = img[0].cpu().float().numpy()
img = (np.transpose(img, (1, 2, 0)) + 1)/2.0 * 255
return img.astype(np.uint8)
def img_save(img, path, name):
img = Image.fromarray(img)
img.save(os.path.join(path, name))
def train(dataset, model, config):
train_loader, val_loader = dataset.train_loader, dataset.val_loader
G_A, G_B, D_A, D_B, cycle_A, cycle_B = [],[],[],[],[],[]
for epoch in range(config.n_epochs):
n_iter = 0
# epoch_start = time.time()
print_iter = random.randint(0,len(train_loader)-1) * config.batch_size
for batch_data in train_loader:
# print(len(batch_data))
A = batch_data[0]
B = batch_data[1]
# iter_start = time.time()
n_iter += config.batch_size
model.Optimize(A, B)
if n_iter % config.print_freq == 0:
print('print loss after %d iter' % (n_iter))
print("G_A: %f, G_B: %f, D_A: %f, D_B: %f, cycle_A: %f, cycle_B: %f" % (model.genLoss_A,model.genLoss_B,model.disLoss_A,model.disLoss_B,model.cycleLoss_A,model.cycleLoss_B))
# if epoch % config.save_freq == 0:
# # print('saving the model, epoch %d' % (epoch))
# # model.model_save(epoch)
# TO DO: print pictures from model.real_A,fake_B,rec_A,real_B,fake_A,rec_B
if n_iter == print_iter:
real_A, real_B = tensor2img(model.real_A), tensor2img(model.real_B)
fake_A, fake_B = tensor2img(model.fake_A), tensor2img(model.fake_B)
rec_A, rec_B = tensor2img(model.rec_A), tensor2img(model.rec_B)
print("Saving pictures at epoch", epoch)
fig = plt.figure()
ax = fig.add_subplot(3, 2, 1)
ax.imshow(real_A)
ax = fig.add_subplot(3, 2, 2)
ax.imshow(real_B)
ax = fig.add_subplot(3, 2, 3)
ax.imshow(fake_A)
ax = fig.add_subplot(3, 2, 4)
ax.imshow(fake_B)
ax = fig.add_subplot(3, 2, 5)
ax.imshow(rec_A)
ax = fig.add_subplot(3, 2, 6)
ax.imshow(rec_B)
plt.show()
img_save(real_A, config.print_dir, str(epoch) + '_real_A.jpg')
img_save(real_B, config.print_dir, str(epoch) + '_real_B.jpg')
img_save(fake_A, config.print_dir, str(epoch) + '_fake_A.jpg')
img_save(fake_B, config.print_dir, str(epoch) + '_fake_B.jpg')
img_save(rec_A, config.print_dir, str(epoch) + '_rec_A.jpg')
img_save(rec_B, config.print_dir, str(epoch) + '_rec_B.jpg')
model.lr_update()
G_A.append(model.genLoss_A)
G_B.append(model.genLoss_A)
D_A.append(model.disLoss_A)
D_B.append(model.disLoss_B)
cycle_A.append(model.cycleLoss_A)
cycle_B.append(model.cycleLoss_B)
plt.title('G_A')
plt.plot(G_A)
plt.show()
plt.title('G_B')
plt.plot(G_B)
plt.show()
plt.title('D_A')
plt.plot(D_A)
plt.show()
plt.title('D_B')
plt.plot(D_B)
plt.show()
plt.title('cycle_A')
plt.plot(cycle_A)
plt.show()
plt.title('cycle_B')
plt.plot(cycle_B)
plt.show()
return
if __name__ == '__main__':
config = Arguments().parse()
dataSet = FaceDataSet(config, batch_size=config.batch_size, dataset_path=config.dataset_path)
trainModel = CycleGAN(config)
if config.is_train:
train(dataSet, trainModel, config)