-
Notifications
You must be signed in to change notification settings - Fork 5
/
losses.py
176 lines (136 loc) · 6.03 KB
/
losses.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
from torch import cat
from torch import tensor
from torch import reshape
from torch.nn import PairwiseDistance
from robust_loss.robust_loss_pytorch.adaptive import AdaptiveVolumeLossFunction
class AdaptiveLoss(object):
def __init__(self, image_shape, image_device, lambda_reconstruction=1):
super(AdaptiveLoss).__init__()
self.av_loss = lambda x: AdaptiveVolumeLossFunction(
image_size=image_shape, device=image_device
).lossfun(x.permute(0, 2, 3, 4, 1))
self.lambda_reconstruction = lambda_reconstruction
def __call__(self, network_outputs):
originals = network_outputs[("org")]
reconstructions = network_outputs[("rec")]
loss_total = None
summaries = {}
adaptive_loss = (
self.av_loss(originals - reconstructions) * self.lambda_reconstruction
).mean()
summaries[("summaries", "scalar", "Adaptive-Loss")] = adaptive_loss
summaries[
("summaries", "scalar", "Lambda-Reconstruction")
] = self.lambda_reconstruction
for key in network_outputs:
if len(key) == 2:
if key[1] == "ql":
if loss_total is None:
loss_total = network_outputs[key]
else:
loss_total += network_outputs[key]
summaries[
(
"summaries",
"scalar",
"L2-VQ_" + str.upper(str(key[0])) + "-Loss",
)
] = network_outputs[key]
loss_total += adaptive_loss
summaries[("summaries", "scalar", "Total_Loss")] = loss_total
return loss_total, summaries
def set_lambda_reconstruction(self, lambda_reconstruction):
self.lambda_reconstruction = lambda_reconstruction
return self.lambda_reconstruction
class BaurLoss(object):
def __init__(self, lambda_reconstruction=1):
super(BaurLoss).__init__()
self.lambda_reconstruction = lambda_reconstruction
self.lambda_gdl = 0
self.l1_loss = lambda x, y: PairwiseDistance(p=1)(
x.view(x.shape[0], -1), y.view(y.shape[0], -1)
).sum()
self.l2_loss = lambda x, y: PairwiseDistance(p=2)(
x.view(x.shape[0], -1), y.view(y.shape[0], -1)
).sum()
def __call__(self, network_outputs):
originals = network_outputs[("org")]
reconstructions = network_outputs[("rec")]
summaries = {}
loss_total = None
l1_reconstruction = (
self.l1_loss(originals, reconstructions) * self.lambda_reconstruction
)
l2_reconstruction = (
self.l2_loss(originals, reconstructions) * self.lambda_reconstruction
)
summaries[("summaries", "scalar", "L1-Reconstruction-Loss")] = l1_reconstruction
summaries[("summaries", "scalar", "L2-Reconstruction-Loss")] = l2_reconstruction
summaries[
("summaries", "scalar", "Lambda-Reconstruction")
] = self.lambda_reconstruction
originals_gradients = self.__image_gradients(originals)
reconstructions_gradients = self.__image_gradients(reconstructions)
l1_gdl = (
self.l1_loss(originals_gradients[0], reconstructions_gradients[0])
+ self.l1_loss(originals_gradients[1], reconstructions_gradients[1])
+ self.l1_loss(originals_gradients[2], reconstructions_gradients[2])
) * self.lambda_gdl
l2_gdl = (
self.l2_loss(originals_gradients[0], reconstructions_gradients[0])
+ self.l2_loss(originals_gradients[1], reconstructions_gradients[1])
+ self.l2_loss(originals_gradients[2], reconstructions_gradients[2])
) * self.lambda_gdl
summaries[("summaries", "scalar", "L1-Image_Gradient-Loss")] = l1_gdl
summaries[("summaries", "scalar", "L2-Image_Gradient-Loss")] = l2_gdl
summaries[("summaries", "scalar", "Lambda-Image_Gradient")] = self.lambda_gdl
for key in network_outputs:
if len(key) == 2:
if key[1] == "ql":
if loss_total is None:
loss_total = network_outputs[key]
else:
loss_total += network_outputs[key]
summaries[
(
"summaries",
"scalar",
"L2-VQ_" + str.upper(str(key[0])) + "-Loss",
)
] = network_outputs[key]
loss_total += l1_reconstruction + l2_reconstruction + l1_gdl + l2_gdl
summaries[("summaries", "scalar", "Total_Loss")] = loss_total
return loss_total, summaries
def set_lambda_reconstruction(self, lambda_reconstruction):
self.lambda_reconstruction = lambda_reconstruction
return self.lambda_reconstruction
def set_lambda_gdl(self, lambda_gdl):
self.lambda_gdl = lambda_gdl
return self.lambda_gdl
@staticmethod
def __image_gradients(image):
input_shape = image.shape
batch_size, features, depth, height, width = input_shape
dz = image[:, :, 1:, :, :] - image[:, :, :-1, :, :]
dy = image[:, :, :, 1:, :] - image[:, :, :, :-1, :]
dx = image[:, :, :, :, 1:] - image[:, :, :, :, :-1]
dzz = tensor(()).new_zeros(
(batch_size, features, 1, height, width),
device=image.device,
dtype=dz.dtype,
)
dz = cat([dz, dzz], 2)
dz = reshape(dz, input_shape)
dyz = tensor(()).new_zeros(
(batch_size, features, depth, 1, width), device=image.device, dtype=dy.dtype
)
dy = cat([dy, dyz], 3)
dy = reshape(dy, input_shape)
dxz = tensor(()).new_zeros(
(batch_size, features, depth, height, 1),
device=image.device,
dtype=dx.dtype,
)
dx = cat([dx, dxz], 4)
dx = reshape(dx, input_shape)
return dx, dy, dz