-
Notifications
You must be signed in to change notification settings - Fork 12
/
utils.py
188 lines (164 loc) · 10.7 KB
/
utils.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
from collections import OrderedDict
import os
import numpy as np
import torch
import torch.nn.functional as F
def change_channel(outputs_norm, scale=1):
row,col,channel=outputs_norm.shape
nx = np.ones((row,col)) -outputs_norm[:, :, 0]
ny = np.ones((row,col)) - outputs_norm[:, :, 1]
nz = outputs_norm[:, :, 2]
new_norm = [nx, nz, ny]
return new_norm
def get_dataList(filename):
f = open(filename, 'r')
data_list = list()
while 1:
line = f.readline()
line = line.strip()
if (not line):
break
data_list.append(line)
f.close()
return data_list
def load_resume_state_dict(model, resume_state_dict):
model_dict = model.state_dict()
# 1. filter out unnecessary keys
resume_state_dict = {k: v for k, v in resume_state_dict.items() if k in model_dict}
# 2. overwrite entries in the existing state dict
model_dict.update(resume_state_dict)
return model_dict
def recursive_glob(rootdir='.', suffix=''):
"""Performs recursive glob with given suffix and rootdir
:param rootdir is the root directory
:param suffix is the suffix to be searched
"""
return [os.path.join(looproot, filename)
for looproot, _, filenames in os.walk(rootdir)
for filename in filenames if filename.endswith(suffix)]
def convert_state_dict(state_dict):
"""Converts a state dict saved from a dataParallel module to normal
module state_dict inplace
:param state_dict is the loaded DataParallel model_state
"""
new_state_dict = OrderedDict()
for k, v in state_dict.items():
name = k[7:] # remove 'module'
new_state_dict[name] = v
return new_state_dict
def norm_tf(outputs):
bz, ch, img_rows, img_cols = outputs.size()
outputs = outputs.permute(0,2,3,1).contiguous().view(-1,ch)
outputs_n = F.normalize(outputs,p=2)
outputs_n = 0.5*(outputs_n+1)
outputs_n = outputs_n.view(-1, img_rows, img_cols, ch)
outputs_n = outputs_n.permute(0,3,1,2)
return outputs_n
def norm_sm(outputs):
bz, ch, img_rows, img_cols = outputs.size()
outputs = outputs.permute(0,2,3,1).contiguous().view(-1,ch)
outputs_n = F.normalize(outputs,p=2)
outputs_n = outputs_n.view(-1, img_rows, img_cols, ch)
outputs_n = outputs_n.permute(0,3,1,2)
return outputs_n
def norm_imsave(outputs):
# outputs_s = np.squeeze(outputs.data.cpu().numpy(), axis=0)
# outputs_s = outputs_s.transpose(1, 2, 0)
# outputs_s = outputs_s.reshape(-1,3)
# outputs_norm = sk.normalize(outputs_s, norm='l2', axis=1)
# outputs_norm = outputs_norm.reshape(orig_size[0], orig_size[1], 3)
# outputs_norm = 0.5*(outputs_norm+1)
bz, ch, img_rows, img_cols = outputs.size()# bz should be one for imsave
outputs = outputs.permute(0,2,3,1).contiguous().view(-1,ch)
outputs_n = F.normalize(outputs,p=2)
outputs_n = 0.5*(outputs_n+1)
outputs_n = outputs_n.view(-1, img_rows, img_cols, ch)
# outputs_n = outputs_n.permute(0,3,1,2)
return outputs_n
def get_fconv_premodel(model_F, resume_state_dict):
model_params = model_F.state_dict()
# copy parameter from resume_state_dict
# conv1, conv+bn+conv+bn
model_params['module.conv1.conv.0.weight'] = resume_state_dict['module.conv1.conv.0.weight']
model_params['module.conv1.conv.0.bias'] = resume_state_dict['module.conv1.conv.0.bias']
model_params['module.conv1.conv.1.weight'] = resume_state_dict['module.conv1.conv.1.weight']
model_params['module.conv1.conv.1.bias'] = resume_state_dict['module.conv1.conv.1.bias']
model_params['module.conv1.conv.3.weight'] = resume_state_dict['module.conv1.conv.3.weight']
model_params['module.conv1.conv.3.bias'] = resume_state_dict['module.conv1.conv.3.bias']
model_params['module.conv1.conv.4.weight'] = resume_state_dict['module.conv1.conv.4.weight']
model_params['module.conv1.conv.4.bias'] = resume_state_dict['module.conv1.conv.4.bias']
# conv2, conv+bn+conv+bn
model_params['module.conv2.conv.0.weight'] = resume_state_dict['module.conv2.conv.0.weight']
model_params['module.conv2.conv.0.bias'] = resume_state_dict['module.conv2.conv.0.bias']
model_params['module.conv2.conv.1.weight'] = resume_state_dict['module.conv2.conv.1.weight']
model_params['module.conv2.conv.1.bias'] = resume_state_dict['module.conv2.conv.1.bias']
model_params['module.conv2.conv.3.weight'] = resume_state_dict['module.conv2.conv.3.weight']
model_params['module.conv2.conv.3.bias'] = resume_state_dict['module.conv2.conv.3.bias']
model_params['module.conv2.conv.4.weight'] = resume_state_dict['module.conv2.conv.4.weight']
model_params['module.conv2.conv.4.bias'] = resume_state_dict['module.conv2.conv.4.bias']
# conv3, conv+bn+conv+bn+conv+bn
model_params['module.conv3.conv.0.weight'] = resume_state_dict['module.conv3.conv.0.weight']
model_params['module.conv3.conv.0.bias'] = resume_state_dict['module.conv3.conv.0.bias']
model_params['module.conv3.conv.1.weight'] = resume_state_dict['module.conv3.conv.1.weight']
model_params['module.conv3.conv.1.bias'] = resume_state_dict['module.conv3.conv.1.bias']
model_params['module.conv3.conv.3.weight'] = resume_state_dict['module.conv3.conv.3.weight']
model_params['module.conv3.conv.3.bias'] = resume_state_dict['module.conv3.conv.3.bias']
model_params['module.conv3.conv.4.weight'] = resume_state_dict['module.conv3.conv.4.weight']
model_params['module.conv3.conv.4.bias'] = resume_state_dict['module.conv3.conv.4.bias']
model_params['module.conv3.conv.6.weight'] = resume_state_dict['module.conv3.conv.6.weight']
model_params['module.conv3.conv.6.bias'] = resume_state_dict['module.conv3.conv.6.bias']
model_params['module.conv3.conv.7.weight'] = resume_state_dict['module.conv3.conv.7.weight']
model_params['module.conv3.conv.7.bias'] = resume_state_dict['module.conv3.conv.7.bias']
# # conv4, conv+bn+conv+bn+conv+bn
# model_params['module.conv4.conv.0.weight'] = resume_state_dict['module.conv4.conv.0.weight']
# model_params['module.conv4.conv.0.bias'] = resume_state_dict['module.conv4.conv.0.bias']
# model_params['module.conv4.conv.1.weight'] = resume_state_dict['module.conv4.conv.1.weight']
# model_params['module.conv4.conv.1.bias'] = resume_state_dict['module.conv4.conv.1.bias']
# model_params['module.conv4.conv.3.weight'] = resume_state_dict['module.conv4.conv.3.weight']
# model_params['module.conv4.conv.3.bias'] = resume_state_dict['module.conv4.conv.3.bias']
# model_params['module.conv4.conv.4.weight'] = resume_state_dict['module.conv4.conv.4.weight']
# model_params['module.conv4.conv.4.bias'] = resume_state_dict['module.conv4.conv.4.bias']
# model_params['module.conv4.conv.6.weight'] = resume_state_dict['module.conv4.conv.6.weight']
# model_params['module.conv4.conv.6.bias'] = resume_state_dict['module.conv4.conv.6.bias']
# model_params['module.conv4.conv.7.weight'] = resume_state_dict['module.conv4.conv.7.weight']
# model_params['module.conv4.conv.7.bias'] = resume_state_dict['module.conv4.conv.7.bias']
# # conv5, conv+bn+conv+bn+conv+bn
# model_params['module.conv5.conv.0.weight'] = resume_state_dict['module.conv5.conv.0.weight']
# model_params['module.conv5.conv.0.bias'] = resume_state_dict['module.conv5.conv.0.bias']
# model_params['module.conv5.conv.1.weight'] = resume_state_dict['module.conv5.conv.1.weight']
# model_params['module.conv5.conv.1.bias'] = resume_state_dict['module.conv5.conv.1.bias']
# model_params['module.conv5.conv.3.weight'] = resume_state_dict['module.conv5.conv.3.weight']
# model_params['module.conv5.conv.3.bias'] = resume_state_dict['module.conv5.conv.3.bias']
# model_params['module.conv5.conv.4.weight'] = resume_state_dict['module.conv5.conv.4.weight']
# model_params['module.conv5.conv.4.bias'] = resume_state_dict['module.conv5.conv.4.bias']
# model_params['module.conv5.conv.6.weight'] = resume_state_dict['module.conv5.conv.6.weight']
# model_params['module.conv5.conv.6.bias'] = resume_state_dict['module.conv5.conv.6.bias']
# model_params['module.conv5.conv.7.weight'] = resume_state_dict['module.conv5.conv.7.weight']
# model_params['module.conv5.conv.7.bias'] = resume_state_dict['module.conv5.conv.7.bias']
# # deconv5, conv+bn+conv+bn+conv+bn
# model_params['module.deconv5.conv.0.weight'] = resume_state_dict['module.deconv5.conv.0.weight']
# model_params['module.deconv5.conv.0.bias'] = resume_state_dict['module.deconv5.conv.0.bias']
# model_params['module.deconv5.conv.1.weight'] = resume_state_dict['module.deconv5.conv.1.weight']
# model_params['module.deconv5.conv.1.bias'] = resume_state_dict['module.deconv5.conv.1.bias']
# model_params['module.deconv5.conv.3.weight'] = resume_state_dict['module.deconv5.conv.3.weight']
# model_params['module.deconv5.conv.3.bias'] = resume_state_dict['module.deconv5.conv.3.bias']
# model_params['module.deconv5.conv.4.weight'] = resume_state_dict['module.deconv5.conv.4.weight']
# model_params['module.deconv5.conv.4.bias'] = resume_state_dict['module.deconv5.conv.4.bias']
# model_params['module.deconv5.conv.6.weight'] = resume_state_dict['module.deconv5.conv.6.weight']
# model_params['module.deconv5.conv.6.bias'] = resume_state_dict['module.deconv5.conv.6.bias']
# model_params['module.deconv5.conv.7.weight'] = resume_state_dict['module.deconv5.conv.7.weight']
# model_params['module.deconv5.conv.7.bias'] = resume_state_dict['module.deconv5.conv.7.bias']
# # deconv4, conv+bn+conv+bn+conv+bn
# model_params['module.deconv4.conv.0.weight'] = resume_state_dict['module.deconv4.conv.0.weight']
# model_params['module.deconv4.conv.0.bias'] = resume_state_dict['module.deconv4.conv.0.bias']
# model_params['module.deconv4.conv.1.weight'] = resume_state_dict['module.deconv4.conv.1.weight']
# model_params['module.deconv4.conv.1.bias'] = resume_state_dict['module.deconv4.conv.1.bias']
# model_params['module.deconv4.conv.3.weight'] = resume_state_dict['module.deconv4.conv.3.weight']
# model_params['module.deconv4.conv.3.bias'] = resume_state_dict['module.deconv4.conv.3.bias']
# model_params['module.deconv4.conv.4.weight'] = resume_state_dict['module.deconv4.conv.4.weight']
# model_params['module.deconv4.conv.4.bias'] = resume_state_dict['module.deconv4.conv.4.bias']
# model_params['module.deconv4.conv.6.weight'] = resume_state_dict['module.deconv4.conv.6.weight']
# model_params['module.deconv4.conv.6.bias'] = resume_state_dict['module.deconv4.conv.6.bias']
# model_params['module.deconv4.conv.7.weight'] = resume_state_dict['module.deconv4.conv.7.weight']
# model_params['module.deconv4.conv.7.bias'] = resume_state_dict['module.deconv4.conv.7.bias']
return model_params