-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathautoencoder.py
129 lines (89 loc) · 4.38 KB
/
autoencoder.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
# -*- coding: utf-8 -*-
"""
Created on Wed Oct 11 12:04:33 2023
@author: afandos
"""
import torch
import torch.nn as nn
class ResidualBlock(nn.Module):
def __init__(self, in_channels, out_channels):
super(ResidualBlock, self).__init__()
self.conv1 = nn.Conv2d(in_channels, out_channels, kernel_size=(3,3), stride=(1,1), padding=1)
self.bn1 = nn.BatchNorm2d(out_channels)
self.conv2 = nn.Conv2d(out_channels, out_channels, kernel_size=(3,3), stride=(1,1), padding=1)
self.bn2 = nn.BatchNorm2d(out_channels)
self.relu = nn.ReLU()
def forward(self, x):
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
out += x
out = self.relu(out)
return out
class ResidualBlockTransposed(nn.Module):
def __init__(self, in_channels, out_channels):
super(ResidualBlockTransposed, self).__init__()
self.conv1 = nn.ConvTranspose2d(in_channels, out_channels, kernel_size=(3,3), stride=(1,1), padding=1)
self.bn1 = nn.BatchNorm2d(out_channels)
self.conv2 = nn.ConvTranspose2d(out_channels, out_channels, kernel_size=(3,3), stride=(1,1), padding=1)
self.bn2 = nn.BatchNorm2d(out_channels)
self.relu = nn.ReLU()
def forward(self, x):
out = self.conv1(x)
out = self.bn1(out)
out = self.relu(out)
out = self.conv2(out)
out = self.bn2(out)
out += x
out = self.relu(out)
return out
class AutoEncoder(nn.Module):
"""
This autoencoder creates the normal flow matrix from two frames.
Frames are (3, 480, 640), they are concatenated so that a tensor of size
(6, 480, 640) is obtained, and then features are extracted and
interpolated to get the normal flow (2, 480, 640).
"""
def __init__(self, in_channels=6, out_channels=2):
super(AutoEncoder, self).__init__()
# Encoder
self.conv1 = nn.Conv2d(in_channels, 64, kernel_size=(7,7), stride=(2,2))
self.bn1 = nn.BatchNorm2d(64)
self.relu1 = nn.ReLU()
self.conv2 = nn.Conv2d(64, 128, kernel_size=(5,5), stride=(2,2))
self.bn2 = nn.BatchNorm2d(128)
self.relu2 = nn.ReLU()
self.residualBlock1 = ResidualBlock(128, 128)
self.residualConv1 = nn.Conv2d(128, 256, kernel_size=(3,3), stride=(2,2), padding=1)
self.residualBlock2 = ResidualBlock(256, 256)
self.residualConv2 = nn.Conv2d(256, 512, kernel_size=(3,3), stride=(2,2), padding=1)
# Decoder
self.residualBlockTransposed1 = ResidualBlockTransposed(512, 512)
self.residualConvTransposed1 = nn.ConvTranspose2d(512, 256, kernel_size=(3,3), stride=(2,2), padding=1)
self.residualBlockTransposed2 = ResidualBlockTransposed(256, 256)
self.residualConvTransposed2 = nn.ConvTranspose2d(256, 128, kernel_size=(3,3), stride=(2,2), padding=1)
self.convTransposed1 = nn.ConvTranspose2d(128, 64, kernel_size=(5,5), stride=(2,2))
self.bnTransposed1 = nn.BatchNorm2d(64)
self.reluTransposed1 = nn.ReLU()
self.convTransposed2 = nn.ConvTranspose2d(64, 32, kernel_size=(7,7), stride=(2,2))
self.bnTransposed2 = nn.BatchNorm2d(32)
self.reluTransposed2 = nn.ReLU()
self.convTransposed3 = nn.ConvTranspose2d(32, out_channels, kernel_size=(4,4), stride=(1,1), padding=1)
def forward(self, frame_ant, frame):
x = torch.cat((frame_ant, frame), dim=1)
x = self.relu1(self.bn1(self.conv1(x)))
x = self.relu2(self.bn2(self.conv2(x)))
x = self.residualBlock1(x)
x = self.residualConv1(x)
x = self.residualBlock2(x)
x = self.residualConv2(x)
x = self.residualBlockTransposed1(x)
x = self.residualConvTransposed1(x)
x = self.residualBlockTransposed2(x)
x = self.residualConvTransposed2(x)
x = self.reluTransposed1(self.bnTransposed1(self.convTransposed1(x)))
x = self.reluTransposed2(self.bnTransposed2(self.convTransposed2(x)))
x = self.convTransposed3(x)
return x