Ecoder/models.py at master · kakwok/Ecoder · GitHub

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
from tensorflow.keras.layers import Input, Dense, Conv2D, MaxPooling2D, UpSampling2D
from tensorflow.keras.models import Model
from tensorflow.keras import backend as K
import tensorflow.keras as kr
import qkeras as qkr


def autoCNN(N_filters=[16,8,4],conv_filter=(3,3),pool_filter=(2,2),shape=(12,4,1),weights_f=''):

  input_img = Input(shape)  # adapt this if using `channels_first` image data format

  for i,ifilter in enumerate(N_filters):
    if i==0:
      ### use input_img for first conv layer
      x = Conv2D(ifilter, conv_filter, activation='relu', padding='same')(input_img)
    else:
      x = Conv2D(ifilter, conv_filter, activation='relu', padding='same')(x)
    if i!=len(N_filters)-1:
      x = MaxPooling2D(pool_filter, padding='same')(x)
    else:
      ### make encoded tensor for last pooling layer
      encoded = MaxPooling2D(pool_filter, padding='same')(x)

  ## decoder part
  for i,ifilter in enumerate(N_filters[::-1]):
    if i==0:
      ### use input_img for first conv layer
      x = Conv2D(ifilter, conv_filter, activation='relu', padding='same')(encoded)
    elif i==len(N_filters)-1:
      x = Conv2D(ifilter, conv_filter, activation='relu')(x)        # no padding for last conv layer to keep the shape
    else:
      x = Conv2D(ifilter, conv_filter, activation='relu', padding='same')(x)
    x = UpSampling2D(pool_filter)(x)

  ## output layer
  decoded = Conv2D(1, conv_filter, activation='sigmoid', padding='same')(x)

  encoder     = Model(input_img, encoded)
  autoencoder = Model(input_img, decoded)
  autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')

  if not weights_f=='':
    autoencoder.load_weights(weights_f)
  #autoencoder.summary()

  return autoencoder,encoder

#def deepAuto(encoding_dim = 4,input_dim=48,weights_f=''):
def deepAuto(dims=[],weights_f=''):

  # dims = sequential number of layers up-to the deepest one
  encoding_dim = dims[-1]
  input_dim    = dims[0]

  # "encoded" is the encoded representation of the input
  input_img = Input(shape=(input_dim,))
  encoded = Dense(dims[1], activation='relu')(input_img)
  for d in dims[2:]:
    encoded = Dense(d, activation='relu')(encoded)
  r_dims = dims[::-1]
  decoded = Dense(r_dims[1], activation='relu')(encoded)
  for d in r_dims[2:-1]:
    decoded = Dense(d, activation='relu')(decoded)  ## repeat this line
  decoded = Dense(input_dim, activation='sigmoid')(decoded)

  encoder     = Model(input_img, encoded)
  autoencoder = Model(input_img, decoded)
  autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')
  #autoencoder.summary()

  if not weights_f=='':
    autoencoder.load_weights(weights_f)

  return autoencoder,encoder

def QautoCNN(shape=(12,4,1), weights_f=''):
  input_img = Input(shape=shape,name='input')
  x = Conv2D(16, (3, 3), activation='relu', padding='same',name='conv2d_0_m')(input_img)
  x = MaxPooling2D((2, 2), padding='same',name='mp_0')(x)
  x = Conv2D(8, (3, 3), activation='relu', padding='same',name='conv2d_1_m')(x)
  x = MaxPooling2D((2, 2), padding='same',name='mp_1')(x)
  x = Conv2D(4, (3, 3), activation='relu', padding='same',name='conv2d_2_m')(x)
  encoded = MaxPooling2D((2, 2), padding='same',name='mp_2')(x)

  # at this point the representation is (4, 4, 8) i.e. 128-dimensional

  x = Conv2D(4, (3, 3), activation='relu', padding='same',name='conv2d_3_m')(encoded)
  x = UpSampling2D((2, 2),name='up_0')(x)
  x = Conv2D(8, (3, 3), activation='relu', padding='same',name='conv2d_4_m')(x)
  x = UpSampling2D((2, 2),name='up_1')(x)
  x = Conv2D(16, (3, 3), activation='relu',name='conv2d_5_m')(x)
  x = UpSampling2D((2, 2),name='up_2')(x)
  decoded = Conv2D(1, (3, 3), activation='sigmoid', padding='same',name='conv2d_6_m')(x)

  q_dict = {
      "conv2d_0_m": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "conv2d_1_m": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "conv2d_2_m": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "conv2d_3_m": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "conv2d_4_m": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "conv2d_5_m": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "conv2d_6_m": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "mp_0": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "mp_1": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "mp_2": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "up_0": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "up_1": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      },
      "up_2": {
          "kernel": "quantized_bits(4)",
          "bias": "quantized_bits(4)"
      }
  }

  autoencoder     = kr.Model(input_img, decoded)
  encoder         = kr.Model(input_img, encoded)
  qautoencoder, _ = qkr.model_quantize(autoencoder, q_dict, 4)
  qencoder, _     = qkr.model_quantize(encoder, q_dict, 4)
  qautoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')
  #qautoencoder.summary()

  if not weights_f=='':
    qautoencoder.load_weights(weights_f)

  return qautoencoder,qencoder