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
|
import torch
import torch.nn as nn
import torch.nn.functional as F
from models.layers import VGGConv2d, DCGANConvTranspose2d
class Encoder(nn.Module):
"""Squeeze input feature map to lower dimension"""
def __init__(
self,
in_channels: int = 3,
frame_size: tuple[int, int] = (64, 48),
feature_channels: int = 64,
output_dims: tuple[int, int, int] = (192, 192, 128)
):
super().__init__()
h_0, w_0 = frame_size
h_1, w_1 = h_0 // 2, w_0 // 2
h_2, w_2 = h_1 // 2, w_1 // 2
# Appearance features, canonical features, pose features
(self.f_a_dim, self.f_c_dim, self.f_p_dim) = output_dims
# Conv1 in_channels x H x W
# -> feature_map_size x H x W
self.conv1 = VGGConv2d(in_channels, feature_channels)
# MaxPool1 feature_map_size x H x W
# -> feature_map_size x H//2 x W//2
self.max_pool1 = nn.AdaptiveMaxPool2d((h_1, w_1))
# Conv2 feature_map_size x H//2 x W//2
# -> feature_map_size*4 x H//2 x W//2
self.conv2 = VGGConv2d(feature_channels, feature_channels * 4)
# MaxPool2 feature_map_size*4 x H//2 x W//2
# -> feature_map_size*4 x H//4 x W//4
self.max_pool2 = nn.AdaptiveMaxPool2d((h_2, w_2))
# Conv3 feature_map_size*4 x H//4 x W//4
# -> feature_map_size*8 x H//4 x W//4
self.conv3 = VGGConv2d(feature_channels * 4, feature_channels * 8)
# Conv4 feature_map_size*8 x H//4 x W//4
# -> feature_map_size*8 x H//4 x W//4 (for large dataset)
self.conv4 = VGGConv2d(feature_channels * 8, feature_channels * 8)
def forward(self, x):
x = self.conv1(x)
x = self.max_pool1(x)
x = self.conv2(x)
x = self.max_pool2(x)
x = self.conv3(x)
x = self.conv4(x)
f_appearance, f_canonical, f_pose = x.split(
(self.f_a_dim, self.f_c_dim, self.f_p_dim), dim=1
)
return f_appearance, f_canonical, f_pose
class Decoder(nn.Module):
"""Upscale embedding to original image"""
def __init__(
self,
feature_channels: int = 64,
out_channels: int = 3,
):
super().__init__()
self.feature_channels = feature_channels
# TransConv1 feature_map_size*8 x H x W
# -> feature_map_size*4 x H x W
self.trans_conv1 = DCGANConvTranspose2d(feature_channels * 8,
feature_channels * 4,
kernel_size=3,
stride=1,
padding=1)
# TransConv2 feature_map_size*4 x H x W
# -> feature_map_size*2 x H*2 x W*2
self.trans_conv2 = DCGANConvTranspose2d(feature_channels * 4,
feature_channels * 2)
# TransConv3 feature_map_size*2 x H*2 x W*2
# -> feature_map_size x H*2 x W*2
self.trans_conv3 = DCGANConvTranspose2d(feature_channels * 2,
feature_channels,
kernel_size=3,
stride=1,
padding=1)
# TransConv4 feature_map_size x H*2 x W*2
# -> in_channels x H*4 x W*4
self.trans_conv4 = DCGANConvTranspose2d(feature_channels, out_channels,
is_last_layer=True)
def forward(self, f_appearance, f_canonical, f_pose):
x = torch.cat((f_appearance, f_canonical, f_pose), dim=1)
x = self.trans_conv1(x)
x = self.trans_conv2(x)
x = self.trans_conv3(x)
x = torch.sigmoid(self.trans_conv4(x))
return x
class AutoEncoder(nn.Module):
def __init__(
self,
channels: int = 3,
frame_size: tuple[int, int] = (64, 48),
feature_channels: int = 64,
embedding_dims: tuple[int, int, int] = (192, 192, 128)
):
super().__init__()
self.embedding_dims = embedding_dims
self.encoder = Encoder(channels, frame_size,
feature_channels, embedding_dims)
self.decoder = Decoder(feature_channels, channels)
def forward(self, x_c1_t2, x_c1_t1=None, x_c2_t2=None):
n, t, c, h, w = x_c1_t2.size()
# x_c1_t2 is the frame for later module
x_c1_t2_ = x_c1_t2.view(n * t, c, h, w)
(f_a_c1_t2_, f_c_c1_t2_, f_p_c1_t2_) = self.encoder(x_c1_t2_)
f_size = [torch.Size([n, t, embedding_dim, h // 4, w // 4])
for embedding_dim in self.embedding_dims]
f_a_c1_t2 = f_a_c1_t2_.view(f_size[0])
f_c_c1_t2 = f_c_c1_t2_.view(f_size[1])
f_p_c1_t2 = f_p_c1_t2_.view(f_size[2])
if self.training:
# t1 is random time step, c2 is another condition
x_c1_t1_ = x_c1_t1.view(n * t, c, h, w)
(f_a_c1_t1_, f_c_c1_t1_, _) = self.encoder(x_c1_t1_)
x_c2_t2_ = x_c2_t2.view(n * t, c, h, w)
(_, f_c_c2_t2_, f_p_c2_t2_) = self.encoder(x_c2_t2_)
x_c1_t2_pred_ = self.decoder(f_a_c1_t1_, f_c_c1_t1_, f_p_c1_t2_)
x_c1_t2_pred = x_c1_t2_pred_.view(n, t, c, h, w)
xrecon_loss = torch.stack([
F.mse_loss(x_c1_t2[:, i], x_c1_t2_pred[:, i])
for i in range(t)
]).sum()
f_c_c1_t1 = f_c_c1_t1_.view(f_size[1])
f_c_c2_t2 = f_c_c2_t2_.view(f_size[1])
cano_cons_loss = torch.stack([
F.mse_loss(f_c_c1_t1[:, i], f_c_c1_t2[:, i])
+ F.mse_loss(f_c_c1_t2[:, i], f_c_c2_t2[:, i])
for i in range(t)
]).mean()
f_p_c2_t2 = f_p_c2_t2_.view(f_size[2])
pose_sim_loss = F.mse_loss(f_p_c1_t2.mean(1), f_p_c2_t2.mean(1))
return (
(f_a_c1_t2, f_c_c1_t2, f_p_c1_t2),
(xrecon_loss, cano_cons_loss, pose_sim_loss * 10)
)
else: # evaluating
return f_a_c1_t2, f_c_c1_t2, f_p_c1_t2
|
