forked from jehovahxu/ca-gan
-
Notifications
You must be signed in to change notification settings - Fork 0
/
data.py
204 lines (159 loc) · 6.73 KB
/
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
# -*- coding: utf-8 -*-
# @Author: JacobShi777
import cv2
import os
import random
import numpy as np
import torch
import random
import scipy.io as sio
import cPickle
from torch.autograd import Variable
import torchvision.transforms as transforms
import torch.utils.data as data
def formnames(infofile, if_train):
infofile = infofile[0] if if_train else infofile[1]
res = []
with open(infofile, 'r') as f:
for line in f:
line = line.strip()
res.append(line)
return res
def input_transform(if_train, opt):
if if_train:
transform = transforms.Compose([transforms.ToTensor()])
else:
transform = transforms.Compose([transforms.ToTensor()])
return transform
def target_transform(if_train, opt):
if if_train:
transform = transforms.Compose([transforms.ToTensor()])
else:
transform = transforms.Compose([transforms.ToTensor()])
return transform
def load_inputs(imgpath, matpath, opt, if_train):
# load photos and paraing
imgpath = os.path.join(opt.root, imgpath)
matpath = os.path.join(opt.root, matpath)
img = cv2.imread(imgpath)
# img = img.astype(np.float32)
# img = img / 255
# imgmean = []
# imgstd = []
# for i in range(3):
# imgmean.append(float(img[:,:,i].mean()))
# imgstd.append(float(img[:,:,i].var()))
# transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((\
# imgmean[0], imgmean[1], imgmean[2]),(imgstd[0], imgstd[1], imgstd[2]))])
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])
img = transform(img)
img = np.transpose(img.numpy(), (1, 2, 0))
img_fl = mat_merge(img, matpath)
img_fl = np.transpose(img_fl, (2, 0, 1))
if if_train:
# img_fl = cv2.resize(img_fl, (opt.loadSize, opt.loadSize))
img_fl = zero_padding(img_fl, opt.loadSize, opt.loadSize - img_fl.shape[1], opt.loadSize - img_fl.shape[2])
img_fl = mat_process(img_fl)
else:
# img_fl = cv2.resize(img_fl, (opt.fineSize, opt.fineSize))
img_fl = zero_padding(img_fl, opt.fineSize, opt.fineSize - img_fl.shape[1], opt.fineSize - img_fl.shape[2])
img_fl = mat_process(img_fl)
return img_fl
def load_targets(imgpath, opt, if_train):
# load sketches
imgpath = os.path.join(opt.root, imgpath)
img = cv2.imread(imgpath, cv2.IMREAD_GRAYSCALE)
img = img.astype(np.float32)
img = img / 255
img = img.reshape(1, img.shape[0], img.shape[1])
if if_train:
# img = cv2.resize(img, (opt.loadSize, opt.loadSize))
img = zero_padding(img, opt.loadSize, opt.loadSize - img.shape[1], opt.loadSize - img.shape[2])
else:
# img = cv2.resize(img, (opt.fineSize, opt.fineSize))
img = zero_padding(img, opt.fineSize, opt.fineSize - img.shape[1], opt.fineSize - img.shape[2])
return img
def mat_merge(img, matpath):
facelabel = sio.loadmat(matpath)
temp = facelabel['res_label']
img = np.concatenate((img, temp), axis=2)
return img
def zero_padding(img, size0, pad1, pad2):
zero_padding = np.zeros((img.shape[0], size0, size0), dtype=np.float32)
pad1 = pad1 / 2
pad2 = pad2 / 2
zero_padding[:, pad1:size0 - pad1, pad2:size0 - pad2] = img
return zero_padding
def mat_process(img_fl):
img_fl = img_fl.astype(np.float32)
img = img_fl[0:3, :, :]
temp = img_fl[3:, :, :]
l0 = temp[0, :, :]
l1 = temp[1, :, :]
l2 = temp[2, :, :] + temp[3, :, :]
l2 = np.where(l2 > 1, 1, l2)
l3 = temp[4, :, :] + temp[5, :, :]
l3 = np.where(l3 > 1, 1, l3)
l4 = temp[6, :, :]
l5 = temp[7, :, :] + temp[9, :, :]
l5 = np.where(l5 > 1, 1, l5)
l6 = temp[8, :, :]
l7 = temp[10, :, :]
# merge
img = np.concatenate((img, l0.reshape(1, l0.shape[0], l0.shape[1])), axis=0)
img = np.concatenate((img, l1.reshape(1, l1.shape[0], l1.shape[1])), axis=0)
img = np.concatenate((img, l2.reshape(1, l2.shape[0], l2.shape[1])), axis=0)
img = np.concatenate((img, l3.reshape(1, l3.shape[0], l3.shape[1])), axis=0)
img = np.concatenate((img, l4.reshape(1, l4.shape[0], l4.shape[1])), axis=0)
img = np.concatenate((img, l5.reshape(1, l5.shape[0], l5.shape[1])), axis=0)
img = np.concatenate((img, l6.reshape(1, l6.shape[0], l6.shape[1])), axis=0)
img = np.concatenate((img, l7.reshape(1, l7.shape[0], l7.shape[1])), axis=0)
return img
class DatasetFromFolder(data.Dataset):
def __init__(self, opt, if_train):
super(DatasetFromFolder, self).__init__()
self.if_train = if_train
self.opt = opt
self.imgnames = formnames(opt.infofile, self.if_train)
self.input_transform_train = input_transform(self.if_train, self.opt)
self.target_transform_train = target_transform(self.if_train, self.opt)
def __getitem__(self, index):
imgname = self.imgnames[index]
item = imgname.split('||')
inputs = load_inputs(item[0], item[2], self.opt, self.if_train)
targets = load_targets(item[1], self.opt, self.if_train)
identity = torch.LongTensor([int(item[3])])
if self.if_train:
w_offset = random.randint(0, self.opt.loadSize - self.opt.fineSize - 1)
h_offset = random.randint(0, self.opt.loadSize - self.opt.fineSize - 1)
# w_offset = 0
# h_offset = 0
inputs = inputs[:, h_offset:h_offset + self.opt.fineSize, h_offset:h_offset + self.opt.fineSize]
targets = targets[:, h_offset:h_offset + self.opt.fineSize, h_offset:h_offset + self.opt.fineSize]
location = torch.LongTensor([int(h_offset), int(w_offset)])
else:
location = torch.LongTensor([15, 15])
inputs = torch.from_numpy(inputs)
targets = torch.from_numpy(targets)
return inputs, targets, identity, location
def __len__(self):
return len(self.imgnames)
def checkpaths(opt):
if not os.path.exists(opt.checkpoint):
os.mkdir(opt.checkpoint)
if not os.path.exists(opt.gen_root):
os.mkdir(opt.gen_root)
def checkpoint(epoch, netD, netG, netE):
netD_out_path = "./checkpoint/netD_epoch_{}.weight".format(epoch)
torch.save(netD.state_dict(), netD_out_path)
netG_out_path = "./checkpoint/netG_epoch_{}.weight".format(epoch)
torch.save(netG.state_dict(), netG_out_path)
netE_out_path = "./checkpoint/netE_epoch_{}.weight".format(epoch)
torch.save(netE.state_dict(), netE_out_path)
print("Checkpoint saved to {} and {}".format(netD_out_path, netG_out_path))
def usedtime(strat_time, end_time):
delta = int(end_time - strat_time)
hours = delta // 3600
minutes = (delta - hours * 3600) // 60
seconds = delta - hours * 3600 - minutes * 60
return ('%2d:%2d:%2d' % (hours, minutes, seconds))