-
Notifications
You must be signed in to change notification settings - Fork 21
/
model.py
123 lines (90 loc) · 4.13 KB
/
model.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
from __future__ import print_function
import torch
import torch.nn as nn
import numpy as np
import torch.nn.functional as F
class PoseFeature(nn.Module):
def __init__(self, num_points = 6890):
super(PoseFeature, self).__init__()
self.conv1 = torch.nn.Conv1d(3, 64, 1)
self.conv2 = torch.nn.Conv1d(64, 128, 1)
self.conv3 = torch.nn.Conv1d(128, 1024, 1)
self.norm1 = torch.nn.InstanceNorm1d(64)
self.norm2 = torch.nn.InstanceNorm1d(128)
self.norm3 = torch.nn.InstanceNorm1d(1024)
def forward(self, x):
x = F.relu(self.norm1(self.conv1(x)))
x = F.relu(self.norm2(self.conv2(x)))
x = F.relu(self.norm3(self.conv3(x)))
return x
class SPAdaIN(nn.Module):
def __init__(self,norm,input_nc,planes):
super(SPAdaIN,self).__init__()
self.conv_weight = nn.Conv1d(input_nc, planes, 1)
self.conv_bias = nn.Conv1d(input_nc, planes, 1)
self.norm = norm(planes)
def forward(self,x,addition):
x = self.norm(x)
weight = self.conv_weight(addition)
bias = self.conv_bias(addition)
out = weight * x + bias
return out
class SPAdaINResBlock(nn.Module):
def __init__(self,input_nc,planes,norm=nn.InstanceNorm1d,conv_kernel_size=1,padding=0):
super(SPAdaINResBlock,self).__init__()
self.spadain1 = SPAdaIN(norm=norm,input_nc=input_nc,planes=planes)
self.relu = nn.ReLU()
self.conv1 = nn.Conv1d(planes, planes, kernel_size=conv_kernel_size, stride=1, padding=padding)
self.spadain2 = SPAdaIN(norm=norm,input_nc=input_nc,planes=planes)
self.conv2 = nn.Conv1d(planes,planes,kernel_size=conv_kernel_size, stride=1, padding=padding)
self.spadain_res = SPAdaIN(norm=norm,input_nc=input_nc,planes=planes)
self.conv_res=nn.Conv1d(planes,planes,kernel_size=conv_kernel_size, stride=1, padding=padding)
def forward(self,x,addition):
out = self.spadain1(x,addition)
out = self.relu(out)
out = self.conv1(out)
out = self.spadain2(out,addition)
out = self.relu(out)
out = self.conv2(out)
residual = x
residual = self.spadain_res(residual,addition)
residual = self.relu(residual)
residual = self.conv_res(residual)
out = out + residual
return out
class Decoder(nn.Module):
def __init__(self, bottleneck_size = 1024):
self.bottleneck_size = bottleneck_size
super(Decoder, self).__init__()
self.conv1 = torch.nn.Conv1d(self.bottleneck_size, self.bottleneck_size, 1)
self.conv2 = torch.nn.Conv1d(self.bottleneck_size, self.bottleneck_size//2, 1)
self.conv3 = torch.nn.Conv1d(self.bottleneck_size//2, self.bottleneck_size//4, 1)
self.conv4 = torch.nn.Conv1d(self.bottleneck_size//4, 3, 1)
self.spadain_block1 = SPAdaINResBlock(input_nc=3 ,planes=self.bottleneck_size)
self.spadain_block2 = SPAdaINResBlock(input_nc=3 ,planes=self.bottleneck_size//2)
self.spadain_block3 = SPAdaINResBlock(input_nc=3 ,planes=self.bottleneck_size//4)
self.norm1 = torch.nn.InstanceNorm1d(self.bottleneck_size)
self.norm2 = torch.nn.InstanceNorm1d(self.bottleneck_size//2)
self.norm3 = torch.nn.InstanceNorm1d(self.bottleneck_size//4)
self.th = nn.Tanh()
def forward(self, x, addition):
x = self.conv1(x)
x = self.spadain_block1(x,addition)
x = self.conv2(x)
x = self.spadain_block2(x,addition)
x = self.conv3(x)
x = self.spadain_block3(x,addition)
x = 2*self.th(self.conv4(x))
return x
class NPT(nn.Module):
def __init__(self, num_points = 6890, bottleneck_size = 1024):
super(NPT, self).__init__()
self.num_points = num_points
self.bottleneck_size = bottleneck_size
self.encoder = PoseFeature(num_points = num_points)
self.decoder = Decoder(bottleneck_size = self.bottleneck_size+3)
def forward(self, x1, x2):
x1 = self.encoder(x1)
y = torch.cat((x1, x2), 1)
out =self.decoder(y,x2)
return out.transpose(2,1)