forked from shsun-xq/TreeSegNet
-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathnetdef.py
More file actions
207 lines (170 loc) · 8.27 KB
/
netdef.py
File metadata and controls
207 lines (170 loc) · 8.27 KB
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
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
# coding: utf-8
import logging
logging.basicConfig(level=logging.INFO)
import mxnet as mx
def bottleneck(inputs, k):
x = mx.sym.BatchNorm(data=inputs, momentum=0.99)
x = mx.sym.Activation(data=x, act_type='relu')
x = mx.sym.Convolution(data=x, kernel=(1,1), stride=(1,1), num_filter=k*4)
x = mx.sym.Dropout(x, p=0.2)
return x
def composite_function(inputs, dilate):
x = mx.sym.BatchNorm(data=inputs, momentum=0.99)
x = mx.sym.Activation(data=x, act_type='relu')
x = mx.sym.Convolution(data=x, kernel=(3,3), stride=(1,1), pad=dilate, num_filter=k, dilate=dilate)
x = mx.sym.Dropout(x, p=0.2)
return x
def composite_function_bottleneck(inputs, dilate):
x = bottleneck(inputs, k)
x = composite_function(x, dilate)
return x
def transition(inputs):
x = mx.sym.BatchNorm(data=inputs, momentum=0.99)
x = mx.sym.Convolution(data=x, kernel=(1,1), stride=(1,1), pad=(0,0), num_filter=k)
# x = mx.sym.Dropout(x, p=0.2)
return x
def dense_block(inputs, dilate):
x1 = composite_function(inputs, dilate)
x2 = composite_function(mx.sym.concat(inputs, x1, dim=1), dilate)
x3 = composite_function(mx.sym.concat(inputs, x1, x2, dim=1), dilate)
x4 = composite_function(mx.sym.concat(inputs, x1, x2, x3, dim=1), dilate)
return mx.sym.concat(x1, x2, x3, x4, dim=1)
def conv(data, kernel=(3, 3), stride=(1, 1), pad=(0, 0), num_filter=None, name=None):
return mx.sym.Convolution(data=data, kernel=kernel, stride=stride, pad=pad, num_filter=num_filter, name='conv_{}'.format(name))
def bn_relu(data, name):
return mx.sym.Activation(data=mx.sym.BatchNorm(data=data, momentum=0.99, name='bn_{}'.format(name)), act_type='relu', name='relu_{}'.format(name))
def conv_bn_relu(data, kernel=(3, 3), stride=(1, 1), pad=(0, 0), num_filter=None, name=None):
return bn_relu(conv(data, kernel, stride, pad, num_filter, 'conv_{}'.format(name)), 'relu_{}'.format(name))
def down_block(data, f, name):
x = mx.sym.Pooling(data=data, kernel=(2,2), stride=(2,2), pool_type='max')
# temp = conv_bn_relu(data, (3, 3), (2, 2), (1, 1),
# f, 'layer1_{}'.format(name))
x = conv_bn_relu(x, (3, 3), (1, 1), (1, 1),
f, 'layer2_{}'.format(name))
bn = mx.sym.BatchNorm(data=conv(x, (3, 3), (1, 1), (1, 1), f, 'layer3_{}'.format(
name)), momentum=0.99, name='layer3_bn_{}'.format(name))
bn = bn + x
act = mx.sym.Activation(data=bn, act_type='relu',
name='layer3_relu_{}'.format(name))
return bn, act
def up_block(act, bn, f, p, name):
x = mx.sym.UpSampling(
act, num_filter=p, scale=2, sample_type='bilinear', name='upsample_{}'.format(name))
# temp = mx.sym.Deconvolution(data=act, kernel=(3, 3), stride=(2, 2), pad=(
# 1, 1), adj=(1, 1), num_filter=32, name='layer1_dconv_{}'.format(name))
x = mx.sym.concat(bn, x, dim=1)
x = conv_bn_relu(x, (1,1), (1,1), (0,0), f, 'layer_1x1_{}'.format(name))
temp = conv_bn_relu(x, (3, 3), (1, 1), (1, 1),
f, 'layer2_{}'.format(name))
bn = mx.sym.BatchNorm(data=conv(temp, (3, 3), (1, 1), (1, 1), f, 'layer3_{}'.format(
name)), momentum=0.99, name='layer3_bn_{}'.format(name))
bn = bn + x
return mx.sym.Activation(data=bn, act_type='relu', name='layer3_relu_{}'.format(name))
k = 2
def getNet(n):
global k
k = n
data = mx.sym.Variable('data')
global rawData
rawData = data
x = conv_bn_relu(data, (3, 3), (1, 1), (1, 1), 64, 'conv0_1')
net = conv_bn_relu(x, (3, 3), (1, 1), (1, 1), 64, 'conv0_2')
bn1 = mx.sym.BatchNorm(data=conv(
net, (3, 3), (1, 1), (1, 1), 64, 'conv0_3'), momentum=0.99, name='conv0_3_bn')
bn1 = bn1 + x
act1 = mx.sym.Activation(data=bn1, act_type='relu', name='conv0_3_relu')
global ACT1
# ACT1 = resnextBlock(act1,16,(1,1),False,getLayerName('short'),4)
ACT1 = act1
bn2, act2 = down_block(act1, 128, 'down1')
bn3, act3 = down_block(act2, 256, 'down2')
bn4, act4 = down_block(act3, 512, 'down3')
bn5, act5 = down_block(act4, 512, 'down4')
bn6, act6 = down_block(act5, 512, 'down5')
bn7, act7 = down_block(act6, 512, 'down6')
temp = up_block(act7, bn6, 512, 512, 'up6')
temp = up_block(temp, bn5, 512, 512, 'up5')
temp = up_block(temp, bn4, 256, 512, 'up4')
temp = up_block(temp, bn3, 128, 256, 'up3')
temp = up_block(temp, bn2, 64, 128, 'up2')
temp = up_block(temp, bn1, 32, 64, 'up1')
score1 = conv(temp, (1, 1), (1, 1), (0, 0), 6, 'score1')
net1 = mx.sym.SoftmaxOutput(score1, multi_output=True, name='softmax1')
from yllab import load_data
net1 = confusionTree(inputt=temp,tree=load_data('confusionTree'))
return net1
__NAME_COUNT__ = {}
def getLayerName(name="None"):
n = __NAME_COUNT__.get(name,0)
n = n + 1
__NAME_COUNT__[name] = n
return name+'_%s'%n
def resnextBlock(data,num_filter, stride, dim_match, name,
num_group=32, bn_mom=0.9, workspace=256,):
conv1 = mx.sym.Convolution(data=data, num_filter=int(num_filter*0.5), kernel=(1,1), stride=(1,1), pad=(0,0),
no_bias=True, workspace=workspace, name=name + '_conv1')
bn1 = mx.sym.BatchNorm(data=conv1, fix_gamma=False, eps=2e-5, momentum=bn_mom, name=name + '_bn1')
act1 = mx.sym.Activation(data=bn1, act_type='relu', name=name + '_relu1')
conv2 = mx.sym.Convolution(data=act1, num_filter=int(num_filter*0.5), num_group=num_group, kernel=(3,3), stride=stride, pad=(1,1),
no_bias=True, workspace=workspace, name=name + '_conv2')
bn2 = mx.sym.BatchNorm(data=conv2, fix_gamma=False, eps=2e-5, momentum=bn_mom, name=name + '_bn2')
act2 = mx.sym.Activation(data=bn2, act_type='relu', name=name + '_relu2')
conv3 = mx.sym.Convolution(data=act2, num_filter=num_filter, kernel=(1,1), stride=(1,1), pad=(0,0), no_bias=True,
workspace=workspace, name=name + '_conv3')
bn3 = mx.sym.BatchNorm(data=conv3, fix_gamma=False, eps=2e-5, momentum=bn_mom, name=name + '_bn3')
if dim_match:
shortcut = data
else:
shortcut_conv = mx.sym.Convolution(data=data, num_filter=num_filter, kernel=(1,1), stride=stride, no_bias=True,
workspace=workspace, name=name+'_sc')
shortcut = mx.sym.BatchNorm(data=shortcut_conv, fix_gamma=False, eps=2e-5, momentum=bn_mom, name=name + '_sc_bn')
eltwise = bn3 + shortcut
return mx.sym.Activation(data=eltwise, act_type='relu', name=name + '_relu')
def pipe(data,filters):
if filters == 1:
out = conv_bn_relu(data, (3, 3), (1, 1), (1, 1),
1, getLayerName('conv_bn_relu'))
return out
fs = filters*4 #8--->16
global ACT1
data = mx.sym.concat(data,ACT1, dim=1)
out = resnextBlock(data,fs,(1,1),False,getLayerName('resNext'),min(fs//4,32))
return out
def pipe2(inp,filters):
if filters == 1:
out = conv_bn_relu(inp, (3, 3), (1, 1), (1, 1),
1, getLayerName('conv_bn_relu'))
return out
layer1,layer2 = 5,3
out1 = conv_bn_relu(inp, (3, 3), (1, 1), (1, 1),
filters*layer1, getLayerName('conv_bn_relu'))
out = conv_bn_relu(out1, (3, 3), (1, 1), (1, 1),
filters*layer2, getLayerName('conv_bn_relu'))
# out = conv(inp, (3, 3), (1, 1), (1, 1),
# filters, getLayerName('conv_bn_relu'))
out = mx.sym.concat(inp,out, dim=1)
return out
def confusionTree(inputt=None,tree=None):
if inputt is None:
inputt = mx.sym.Variable('data')
classn = sum(map(len,tree.keys()))
probs = [0]*classn
def walkTree(inp, tree, key):
out = pipe(inp,len(key))
if len(key) == 1:
probs[key[0]]=out
else:
for k,v in tree.items():
walkTree(out,v,k)
walkTree(inputt,tree,tuple(range(classn)))
out = mx.sym.concat(*probs, dim=1)
net = mx.sym.SoftmaxOutput(out, multi_output=True, name='softmax1')
return net
if __name__ == '__main__':
pass
# from yllab import *
net = getNet(6)
# net = confusionTree(tree=tre)
mx.viz.plot_network(net, save_format='pdf', shape={
'data': (1, 3, 640, 640),
'softmax1_label': (1, 640, 640), }).render('TresegNet-short')