-
Notifications
You must be signed in to change notification settings - Fork 1
/
Fast_SCNN.py
162 lines (114 loc) · 6.33 KB
/
Fast_SCNN.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
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
import numpy as np
import tensorflow as tf
def down_sample(input_layer):
ds_layer = tf.keras.layers.Conv2D(32, (3,3), padding='same', strides = (2,2))(input_layer)
ds_layer = tf.keras.layers.BatchNormalization()(ds_layer)
ds_layer = tf.keras.activations.relu(ds_layer)
ds_layer = tf.keras.layers.SeparableConv2D(48, (3,3), padding='same', strides = (2,2))(ds_layer)
ds_layer = tf.keras.layers.BatchNormalization()(ds_layer)
ds_layer = tf.keras.activations.relu(ds_layer)
ds_layer = tf.keras.layers.SeparableConv2D(64, (3,3), padding='same', strides = (2,2))(ds_layer)
ds_layer = tf.keras.layers.BatchNormalization()(ds_layer)
ds_layer = tf.keras.activations.relu(ds_layer)
return ds_layer
def _res_bottleneck(inputs, filters, kernel, t, s, r=False):
tchannel = tf.keras.backend.int_shape(inputs)[-1] * t
#x = conv_block(inputs, 'conv', tchannel, (1, 1), strides=(1, 1))
x = tf.keras.layers.Conv2D(tchannel, (1,1), padding='same', strides = (1,1))(inputs)
x = tf.keras.layers.BatchNormalization()(x)
x = tf.keras.activations.relu(x)
x = tf.keras.layers.DepthwiseConv2D(kernel, strides=(s, s), depth_multiplier=1, padding='same')(x)
x = tf.keras.layers.BatchNormalization()(x)
x = tf.keras.activations.relu(x)
#x = #conv_block(x, 'conv', filters, (1, 1), strides=(1, 1), padding='same', relu=False)
x = tf.keras.layers.Conv2D(filters, (1,1), padding='same', strides = (1,1))(x)
x = tf.keras.layers.BatchNormalization()(x)
if r:
x = tf.keras.layers.add([x, inputs])
return x
"""#### Bottleneck custom method"""
def bottleneck_block(inputs, filters, kernel, t, strides, n):
x = _res_bottleneck(inputs, filters, kernel, t, strides)
for i in range(1, n):
x = _res_bottleneck(x, filters, kernel, t, 1, True)
return x
def global_feature_extractor(lds_layer):
gfe_layer = bottleneck_block(lds_layer, 64, (3, 3), t=6, strides=2, n=3)
print("gfe_layer.shape:", gfe_layer.shape)
gfe_layer = bottleneck_block(gfe_layer, 96, (3, 3), t=6, strides=2, n=3)
print("gfe_layer.shape:", gfe_layer.shape)
gfe_layer = bottleneck_block(gfe_layer, 128, (3, 3), t=6, strides=1, n=3)
print("gfe_layer.shape:", gfe_layer.shape)
gfe_layer = pyramid_pooling_block(gfe_layer, [2,4,6,8], gfe_layer.shape[1], gfe_layer.shape[2])
print("gfe_layer.shape:", gfe_layer.shape)
return gfe_layer
def pyramid_pooling_block(input_tensor, bin_sizes, w, h):
print(w, h)
concat_list = [input_tensor]
#w = 16 # 64
#h = 16 #32
for bin_size in bin_sizes:
x = tf.keras.layers.AveragePooling2D(pool_size=(w//bin_size, h//bin_size),
strides=(w//bin_size, h//bin_size))(input_tensor)
x = tf.keras.layers.Conv2D(128, 3, 2, padding='same')(x)
x = tf.keras.layers.Lambda(lambda x: tf.image.resize(x, (w,h)))(x)
print("x in paramid.shape", x.shape)
concat_list.append(x)
return tf.keras.layers.concatenate(concat_list)
def feature_fusion(lds_layer, gfe_layer):
ff_layer1 = tf.keras.layers.Conv2D(128, (1,1), padding='same', strides = (1,1))(lds_layer)
ff_layer1 = tf.keras.layers.BatchNormalization()(ff_layer1)
#ff_layer1 = tf.keras.activations.relu(ff_layer1)
print("ff_layer1.shape", ff_layer1.shape)
#ss = conv_block(gfe_layer, 'conv', 128, (1,1), padding='same', strides= (1,1), relu=False)
#print(ss.shape, ff_layer1.shape)
ff_layer2 = tf.keras.layers.UpSampling2D((4, 4))(gfe_layer)
print("ff_layer2.shape", ff_layer2.shape)
ff_layer2 = tf.keras.layers.DepthwiseConv2D(128, strides=(1, 1), depth_multiplier=1, padding='same')(ff_layer2)
print("ff_layer2.shape", ff_layer2.shape)
ff_layer2 = tf.keras.layers.BatchNormalization()(ff_layer2)
ff_layer2 = tf.keras.activations.relu(ff_layer2)
ff_layer2 = tf.keras.layers.Conv2D(128, 1, 1, padding='same', activation=None)(ff_layer2)
print("ff_layer2.shape", ff_layer2.shape)
ff_final = tf.keras.layers.add([ff_layer1, ff_layer2])
ff_final = tf.keras.layers.BatchNormalization()(ff_final)
ff_final = tf.keras.activations.relu(ff_final)
print("ff_final.shape", ff_final.shape)
return ff_final
def classifier_layer(ff_final, num_classes):
classifier = tf.keras.layers.SeparableConv2D(128, (3, 3), padding='same',
strides = (1, 1), name = 'DSConv1_classifier')(ff_final)
classifier = tf.keras.layers.BatchNormalization()(classifier)
classifier = tf.keras.activations.relu(classifier)
print("classifier.shape", classifier.shape)
classifier = tf.keras.layers.SeparableConv2D(128, (3, 3), padding='same',
strides = (1, 1), name = 'DSConv2_classifier')(classifier)
classifier = tf.keras.layers.BatchNormalization()(classifier)
classifier = tf.keras.activations.relu(classifier)
print("classifier.shape", classifier.shape)
#change 19 to 20
#classifier = conv_block(classifier, 'conv', 20, (1, 1), strides=(1, 1), padding='same', relu=True)
classifier = tf.keras.layers.Conv2D(num_classes, (1,1), padding='same', strides = (1,1))(classifier)
classifier = tf.keras.layers.BatchNormalization()(classifier)
classifier = tf.keras.activations.relu(classifier)
print("classifier.shape", classifier.shape)
classifier = tf.keras.layers.Dropout(0.3)(classifier)
print("classifier before upsampling:", classifier.shape)
classifier = tf.keras.activations.softmax(classifier)
return classifier
def get_fast_scnn(w, h, num_classes):
"""
input image: (w, h)
"""
input_layer = tf.keras.layers.Input(shape=(w, h, 3), name = 'input_layer')
ds_layer = down_sample(input_layer)
gfe_layer = global_feature_extractor(ds_layer)
ff_final = feature_fusion(ds_layer, gfe_layer)
classifier = classifier_layer(ff_final, num_classes)
fast_scnn = tf.keras.Model(inputs = input_layer , outputs = classifier, name = 'Fast_SCNN')
optimizer = tf.keras.optimizers.SGD(momentum=0.9, lr=0.045)
fast_scnn.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])
return fast_scnn
fast_scnn = get_fast_scnn(512, 512, 2)
fast_scnn.summary()
tf.keras.utils.plot_model(fast_scnn, show_layer_names=True, show_shapes=True)