-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathgionet.py
117 lines (95 loc) · 3.22 KB
/
gionet.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
from math import ceil
import paddle
from paddle.nn.functional import interpolate
from resnet_dilated import ResNet101
from paddle.nn import Conv2D, BatchNorm, Layer, Sequential, Conv1D, ReLU
print(paddle.__version__)
#paddle.disable_static()
class GCNModule(Layer):
def __init__(self, num_channels, num_nodes):
super(GCNModule, self).__init__()
self.conv1 = Conv1D(num_nodes, num_nodes, kernel_size=1)
self.relu = ReLU()
self.conv2 = Conv1D(num_channels, num_channels, kernel_size=1)
def forward(self, inputs):
# inputs.shape(B,C,N）
#print(inputs.shape)
#x=fluid.layers.transpose(inputs, perm=(0, 2, 1))
x=paddle.transpose(inputs, perm=[0, 2, 1])
#print(x.shape)
x = self.conv1(x)
#print(x.shape)
#x = fluid.layers.transpose(x, perm=(0,2,1))
x = paddle.transpose(x, perm=[0,2,1])
#print(x.shape)
x = x - inputs
#print(x.shape)
x = self.conv2(x)
#print(x.shape)
x = self.relu(x)
#print(x.shape)
return x
class GIoNet(Layer):
def __init__(self, num_classes=59, IMG_SIZE = None,backbone='resnet50'):
super(GIoNet, self).__init__()
res = ResNet101(pretrained=False)
# stem: res.conv, res.pool2d_max
self.layer0 = Sequential(
res.conv,
res.pool2d_max
)
self.layer1 = res.layer1
self.layer2 = res.layer2
self.layer3 = res.layer3
self.layer4 = res.layer4
in_channels = 2048
out_channels = 256
# gionet channels 2048-> 256 num_nodes = H' *W'
self.layer5 = Sequential(
Conv2D(in_channels=in_channels, out_channels=out_channels, kernel_size=1),
BatchNorm(out_channels, act='relu')
)
num_node = ceil(IMG_SIZE[0]/8)*ceil(IMG_SIZE[1]/8)
self.gcnmodule = GCNModule(num_channels=out_channels, num_nodes=num_node)
self.classifier = Conv2D(in_channels=out_channels, out_channels=num_classes, kernel_size=1)
#self.pspmoduls = PSPModule(num_channels, [1,2,3,6])
#num_channels *= 2
def forward(self, inputs):
x = self.layer0(inputs)
#print(x.shape)
x = self.layer1(x)
#print(x.shape)
x = self.layer2(x)
#print(x.shape)
x = self.layer3(x)
#print(x.shape)
x = self.layer4(x)
#print(x.shape)
x = self.layer5(x)
#print(x.shape)
n,c,h,w = x.shape
#print(x.shape)
x = paddle.reshape(x, shape=[-1, c, h*w])
#print(x.shape)
x = self.gcnmodule(x)
#print(x.shape)
x = paddle.reshape(x, shape=[-1,c,h,w])
#print(x.shape)
x = self.classifier(x)
#print(x.shape)
x = interpolate(x, inputs.shape[2::],align_corners=False) #2::指 2和3
#print(x.shape)
return x
# aux: tmp_x = layer3
def main():
#x = paddle.tensor.rand((2, 3, 520, 520),dtype='float32')
#print(x.shape)
x = (2, 3, 520, 520)
model = GIoNet(num_classes=59,IMG_SIZE=x[2::])
params_info = paddle.summary(model, x)
print(params_info)
#model.train()
#pred = model(x)
#print(pred.shape)
if __name__ == "__main__":
main()