-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathunaligned_dataset.py
181 lines (149 loc) · 6.34 KB
/
unaligned_dataset.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
import os.path
from data.base_dataset import BaseDataset, get_transform, get_transform2
from data.image_folder import make_dataset
from PIL import Image
import torch
import numpy as np
import random
import scipy.io as sio
class UnalignedDataset_test(BaseDataset):
def initialize(self, opt):
self.opt = opt
self.root = opt.dataroot
self.dir_A = os.path.join(opt.dataroot, opt.phase + 'A')
self.dir_B = os.path.join(opt.dataroot, opt.phase + 'B')
self.A_paths = make_dataset(self.dir_A)
self.B_paths = make_dataset(self.dir_B)
self.A_paths = sorted(self.A_paths)
self.B_paths = sorted(self.B_paths)
self.A_size = len(self.A_paths)
self.B_size = len(self.B_paths)
self.transform = get_transform(opt)
self.transformless = get_transform2(opt)
def __getitem__(self, index):
A_path = self.A_paths[index % self.A_size]
if self.opt.serial_batches:
index_B = index % self.B_size
else:
index_B = random.randint(0, self.B_size - 1)
B_path = self.B_paths[index_B]
# print('(A, B) = (%d, %d)' % (index_A, index_B))
A_img = Image.open(A_path).convert('RGB')
B_img = Image.open(B_path).convert('RGB')
A_img = np.transpose(np.array(A_img), (2, 0, 1))
B_img = np.transpose(np.array(B_img), (2, 0, 1))
A = torch.FloatTensor(A_img) / 255.0
B = torch.FloatTensor(B_img) / 255.0
# Comment out this 2 lines if you want to train G_Dec
A = self.transformless(A)
B = self.transformless(B)
if self.opt.which_direction == 'BtoA':
input_nc = self.opt.output_nc
output_nc = self.opt.input_nc
else:
input_nc = self.opt.input_nc
output_nc = self.opt.output_nc
if input_nc == 1: # RGB to gray
tmp = A[0, ...] * 0.299 + A[1, ...] * 0.587 + A[2, ...] * 0.114
A = tmp.unsqueeze(0)
if output_nc == 1: # RGB to gray
tmp = B[0, ...] * 0.299 + B[1, ...] * 0.587 + B[2, ...] * 0.114
B = tmp.unsqueeze(0)
# X = np.zeros([2,A.shape[1],A.shape[2]])
# X[0,:,:] = C
# X[1,:,:] = D
# X = torch.FloatTensor(X)
return {'A': A, 'B': B, # 'X':X, #'E':E,
'A_paths': A_path, 'B_paths': B_path}
def __len__(self):
return max(self.A_size, self.B_size)
def name(self):
return 'UnalignedDataset'
class UnalignedDataset_Dec(BaseDataset):
def initialize(self, opt):
self.opt = opt
self.root = opt.dataroot
self.dir_A = os.path.join(opt.dataroot, opt.phase + 'A')
self.dir_B = os.path.join(opt.dataroot, opt.phase + 'B')
self.dir_C = os.path.join(opt.dataroot, opt.phase + 'C')
self.dir_D = os.path.join(opt.dataroot, opt.phase + 'D')
self.dir_E = os.path.join(opt.dataroot, opt.phase + 'E')
self.A_paths = make_dataset(self.dir_A)
self.B_paths = make_dataset(self.dir_B)
self.C_paths = make_dataset(self.dir_C)
self.D_paths = make_dataset(self.dir_D)
self.E_paths = make_dataset(self.dir_E)
self.A_paths = sorted(self.A_paths)
self.B_paths = sorted(self.B_paths)
self.C_paths = sorted(self.C_paths)
self.D_paths = sorted(self.D_paths)
self.E_paths = sorted(self.E_paths)
self.A_size = len(self.A_paths)
self.B_size = len(self.B_paths)
self.C_size = len(self.C_paths)
self.D_size = len(self.D_paths)
self.E_size = len(self.E_paths)
self.transform = get_transform(opt)
self.transformless = get_transform2(opt)
def __getitem__(self, index):
# A_path = self.A_paths[index % self.A_size]
B_path = self.B_paths[index % self.B_size]
C_path = self.C_paths[index % self.C_size]
D_path = self.D_paths[index % self.D_size]
E_path = self.E_paths[index % self.E_size]
if self.opt.serial_batches:
index_A = index % self.A_size
else:
index_A = random.randint(0, self.A_size - 1)
A_path = self.A_paths[index_A]
# print('(A, B) = (%d, %d)' % (index_A, index_B))
A_img = Image.open(A_path).convert('RGB')
B_img = Image.open(B_path).convert('RGB')
C_img = Image.open(C_path).convert('RGB')
D_img = Image.open(D_path).convert('RGB')
E_img = Image.open(E_path).convert('RGB')
A_img = np.transpose(np.array(A_img), (2, 0, 1))
B_img = np.transpose(np.array(B_img), (2, 0, 1))
C_img = np.transpose(np.array(C_img), (2, 0, 1))
D_img = np.transpose(np.array(D_img), (2, 0, 1))
E_img = np.transpose(np.array(E_img), (2, 0, 1))
A = torch.FloatTensor(A_img) / 255.0
B = torch.FloatTensor(B_img) / 255.0
C = torch.FloatTensor(C_img) / 255.0
D = torch.FloatTensor(D_img) / 255.0
E = torch.FloatTensor(E_img) / 255.0
# Comment out this 5 lines if you want to train G_Dec
A = self.transformless(A)
B = self.transformless(B)
C = self.transformless(C)
D = self.transformless(D)
E = self.transformless(E)
if self.opt.which_direction == 'BtoA':
input_nc = self.opt.output_nc
output_nc = self.opt.input_nc
else:
input_nc = self.opt.input_nc
output_nc = self.opt.output_nc
if input_nc == 1: # RGB to gray
tmp = A[0, ...] * 0.299 + A[1, ...] * 0.587 + A[2, ...] * 0.114
A = tmp.unsqueeze(0)
if output_nc > 0: # RGB to gray
tmp = B[0, ...] * 0.299 + B[1, ...] * 0.587 + B[2, ...] * 0.114
B = tmp.unsqueeze(0)
tmp = C[0, ...] * 0.299 + C[1, ...] * 0.587 + C[2, ...] * 0.114
C = tmp.unsqueeze(0)
tmp = D[0, ...] * 0.299 + D[1, ...] * 0.587 + D[2, ...] * 0.114
D = tmp.unsqueeze(0)
tmp = E[0, ...] * 0.299 + E[1, ...] * 0.587 + E[2, ...] * 0.114
E = tmp.unsqueeze(0)
X = np.zeros([3, A.shape[1], A.shape[2]])
X[0, :, :] = C
X[1, :, :] = D
X[2, :, :] = E
X = torch.FloatTensor(X)
return {'A': A, 'B': B, 'X': X, 'E': B,
'A_paths': A_path, 'B_paths': B_path}
def __len__(self):
return max(self.A_size, self.B_size)
def name(self):
return 'UnalignedDataset'