guided-infant-generation/src/models/bibranch/G1.py at master · vrai-group/guided-infant-generation · 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
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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
import numpy as np
import tensorflow as tf
from tensorflow.keras import Model
from tensorflow.keras.layers import *
from tensorflow.keras.optimizers import Adam

from models.Model_template import Model_Template

class G1(Model_Template):

    def __init__(self):
        self.architecture = "bibranch"
        self.input_shape = [96, 128, 15]
        self.output_channels = 1
        self.activation_fn = 'relu'
        self.lr_initial_G1 = 2e-5
        super().__init__() # eredita self.model e self.opt

    def _build_model(self):
        inputs = Input(shape = self.input_shape)
        ## Encoder
        # Blocco 1
        Enc_1 = Conv2D(filters=128, kernel_size=3, strides=1, padding='same')(inputs)
        Enc_1 = Activation(self.activation_fn)(Enc_1)

        branch_1_Enc_1 = Conv2D(filters=64, kernel_size=3, strides=1, padding='same')(Enc_1)
        branch_1_Enc_1 = Activation(self.activation_fn)(branch_1_Enc_1)

        branch_1_Enc_1 = Conv2D(filters=64, kernel_size=3, strides=1, padding='same')(branch_1_Enc_1)
        branch_1_Enc_1 = Activation(self.activation_fn)(branch_1_Enc_1)

        branch_2_Enc_1 = Conv2D(filters=64, kernel_size=3, strides=1, padding='same')(Enc_1)
        branch_2_Enc_1 = Activation(self.activation_fn)(branch_2_Enc_1)

        branch_2_Enc_1 = Conv2D(filters=64, kernel_size=3, strides=1, padding='same')(branch_2_Enc_1)
        branch_2_Enc_1 = Activation(self.activation_fn)(branch_2_Enc_1)

        concat_1 = concatenate([branch_1_Enc_1, branch_2_Enc_1])

        # Blocco 2
        Enc_2 = Conv2D(filters=256, kernel_size=2, strides=2)(concat_1)
        Enc_2 = Activation(self.activation_fn)(Enc_2)

        branch_1_Enc_2 = Conv2D(filters=128, kernel_size=3, strides=1, padding='same')(Enc_2)
        branch_1_Enc_2 = Activation(self.activation_fn)(branch_1_Enc_2)

        branch_1_Enc_2 = Conv2D(filters=128, kernel_size=3, strides=1, padding='same')(branch_1_Enc_2)
        branch_1_Enc_2 = Activation(self.activation_fn)(branch_1_Enc_2)

        branch_2_Enc_2 = Conv2D(filters=128, kernel_size=3, strides=1, padding='same')(Enc_2)
        branch_2_Enc_2 = Activation(self.activation_fn)(branch_2_Enc_2)

        branch_2_Enc_2 = Conv2D(filters=128, kernel_size=3, strides=1, padding='same')(branch_2_Enc_2)
        branch_2_Enc_2 = Activation(self.activation_fn)(branch_2_Enc_2)

        concat_2 = concatenate([branch_1_Enc_2, branch_2_Enc_2])

        # Blocco 3
        Enc_3 = Conv2D(filters=384, kernel_size=2, strides=2)(concat_2)
        Enc_3 = Activation(self.activation_fn)(Enc_3)

        branch_1_Enc_3 = Conv2D(filters=192, kernel_size=3, strides=1, padding='same')(Enc_3)
        branch_1_Enc_3 = Activation(self.activation_fn)(branch_1_Enc_3)

        branch_1_Enc_3 = Conv2D(filters=192, kernel_size=3, strides=1, padding='same')(branch_1_Enc_3)
        branch_1_Enc_3 = Activation(self.activation_fn)(branch_1_Enc_3)

        branch_2_Enc_3 = Conv2D(filters=192, kernel_size=3, strides=1, padding='same')(Enc_3)
        branch_2_Enc_3 = Activation(self.activation_fn)(branch_2_Enc_3)

        branch_2_Enc_3 = Conv2D(filters=192, kernel_size=3, strides=1, padding='same')(branch_2_Enc_3)
        branch_2_Enc_3 = Activation(self.activation_fn)(branch_2_Enc_3)

        concat_3 = concatenate([branch_1_Enc_3, branch_2_Enc_3])

        # Blocco 4
        Enc_4 = Conv2D(filters=512, kernel_size=2, strides=2)(concat_3)
        Enc_4 = Activation(self.activation_fn)(Enc_4)

        branch_1_Enc_4 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(Enc_4)
        branch_1_Enc_4 = Activation(self.activation_fn)(branch_1_Enc_4)

        branch_1_Enc_4 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(branch_1_Enc_4)
        branch_1_Enc_4 = Activation(self.activation_fn)(branch_1_Enc_4)

        branch_2_Enc_4 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(Enc_4)
        branch_2_Enc_4 = Activation(self.activation_fn)(branch_2_Enc_4)

        branch_2_Enc_4 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(branch_2_Enc_4)
        branch_2_Enc_4 = Activation(self.activation_fn)(branch_2_Enc_4)

        concat_4 = concatenate([branch_1_Enc_4, branch_2_Enc_4]) #512

        ## Bridge
        # output [num_batch, 98.304] --> 98.304:  12x16x512
        x = Reshape([-1, int(np.prod([12, 16, 512]))])(concat_4)
        # output [num_batch, 64]
        z = Dense(64, activation=None)(x)
        # output [num batch, 24576]
        z = Dense(int(np.prod([12, 16, 128])), activation=None)(z)
        # output [num batch, 12,16,128]
        x = Reshape([12, 16, 128])(z)

        ## Decoder
        # Blocco 1
        long_connection_1 = Concatenate(axis=-1)([x, concat_4]) #512+128=640

        branch_1_Dec_1 = Conv2D(filters=320, kernel_size=3, strides=1, padding='same')(long_connection_1)
        branch_1_Dec_1 = Activation(self.activation_fn)(branch_1_Dec_1)

        branch_1_Dec_1 = Conv2D(filters=320, kernel_size=3, strides=1, padding='same')(
            branch_1_Dec_1)
        branch_1_Dec_1 = Activation(self.activation_fn)(branch_1_Dec_1)

        branch_2_Dec_1 = Conv2D(filters=320, kernel_size=3, strides=1, padding='same')(long_connection_1)
        branch_2_Dec_1 = Activation(self.activation_fn)(branch_2_Dec_1)

        branch_2_Dec_1 = Conv2D(filters=320, kernel_size=3, strides=1, padding='same')(
            branch_2_Dec_1)
        branch_2_Dec_1 = Activation(self.activation_fn)(branch_2_Dec_1)

        Dec_1 = concatenate([branch_1_Dec_1, branch_2_Dec_1]) #640
        Dec_1 = UpSampling2D(size=(2, 2), interpolation="nearest")(Dec_1)
        Dec_1 = Conv2D(filters=384, kernel_size=1, strides=1, padding='same')(Dec_1)

        # Blocco 2
        long_connection_2 = Concatenate(axis=-1)([Dec_1, concat_3]) #384+384=768

        branch_1_Dec_2 = Conv2D(filters=384, kernel_size=3, strides=1, padding='same')(long_connection_2)
        # branch_1_Dec_2 = BatchNormalization()(branch_1_Dec_2)
        branch_1_Dec_2 = Activation(self.activation_fn)(branch_1_Dec_2)

        branch_1_Dec_2 = Conv2D(filters=384, kernel_size=3, strides=1, padding='same')(
            branch_1_Dec_2)
        branch_1_Dec_2 = Activation(self.activation_fn)(branch_1_Dec_2)

        branch_2_Dec_2 = Conv2D(filters=384, kernel_size=3, strides=1, padding='same')(long_connection_2)
        branch_2_Dec_2 = Activation(self.activation_fn)(branch_2_Dec_2)

        branch_2_Dec_2 = Conv2D(filters=384, kernel_size=3, strides=1, padding='same')(
            branch_2_Dec_2)
        branch_2_Dec_2 = Activation(self.activation_fn)(branch_2_Dec_2)

        Dec_2 = concatenate([branch_1_Dec_2, branch_2_Dec_2]) # 768
        Dec_2 = UpSampling2D(size=(2, 2), interpolation="nearest")(Dec_2)
        Dec_2 = Conv2D(filters=256, kernel_size=1, strides=1, padding='same')(Dec_2)

        # Blocco 3
        long_connection_3 = Concatenate(axis=-1)([Dec_2, concat_2])  # 512

        branch_1_Dec_3 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(long_connection_3)
        branch_1_Dec_3 = Activation(self.activation_fn)(branch_1_Dec_3)

        branch_1_Dec_3 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(
            branch_1_Dec_3)
        branch_1_Dec_3 = Activation(self.activation_fn)(branch_1_Dec_3)

        branch_2_Dec_3 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(long_connection_3)
        branch_2_Dec_3 = Activation(self.activation_fn)(branch_2_Dec_3)

        branch_2_Dec_3 = Conv2D(filters=256, kernel_size=3, strides=1, padding='same')(
            branch_2_Dec_3)
        branch_2_Dec_3 = Activation(self.activation_fn)(branch_2_Dec_3)

        Dec_3 = concatenate([branch_1_Dec_3, branch_2_Dec_3])
        Dec_3 = UpSampling2D(size=(2, 2), interpolation="nearest")(Dec_3)
        Dec_3 = Conv2D(filters=128, kernel_size=1, strides=1, padding='same')(Dec_3)

        # Blocco 4
        long_connection_4 = Concatenate(axis=-1)([Dec_3, concat_1])

        branch_1_Dec_4 = Conv2D(filters=128, kernel_size=3, strides=1, padding="same")(long_connection_4)
        branch_1_Dec_4 = Activation(self.activation_fn)(branch_1_Dec_4)

        branch_1_Dec_4 = Conv2D(filters=128, kernel_size=3, strides=1, padding='same')(
            branch_1_Dec_4)
        branch_1_Dec_4 = Activation(self.activation_fn)(branch_1_Dec_4)

        branch_2_Dec_4 = Conv2D(filters=128, kernel_size=3, strides=1, padding="same")(long_connection_4)
        branch_2_Dec_4 = Activation(self.activation_fn)(branch_2_Dec_4)

        branch_2_Dec_4 = Conv2D(filters=128, kernel_size=3, strides=1, padding='same')(
            branch_2_Dec_4)
        branch_2_Dec_4 = Activation(self.activation_fn)(branch_2_Dec_4)

        Dec_4 = concatenate([branch_1_Dec_4, branch_2_Dec_4])  # 256

        outputs = Conv2D(1, 1, 1, padding='same', activation=None)(Dec_4)

        return Model(inputs, outputs)

    def _optimizer(self):
        return Adam(learning_rate=self.lr_initial_G1, beta_1=0.5, beta_2=0.999)

    def prediction(self, Ic, Pt):
        input_G1 = tf.concat([Ic, Pt], axis=-1)
        output_G1 = self.model(input_G1)  # [batch, 96, 128, 1] dtype=float32
        output_G1 = tf.cast(output_G1, dtype=tf.float16)
        return output_G1

    # LOSS
    def PoseMaskloss(self, I_PT1, It, Mt):
        I_PT1 = tf.cast(I_PT1, dtype=tf.float32)
        It = tf.cast(It, dtype=tf.float32)
        Mt = tf.cast(Mt, dtype=tf.float32)

        primo_membro = tf.reduce_mean(tf.abs(I_PT1 - It))  # L1 loss
        secondo_membro = tf.reduce_mean(tf.abs(I_PT1 - It) * Mt)
        PoseMaskLoss1 = primo_membro + secondo_membro

        return PoseMaskLoss1

    # METRICHE
    def ssim(self, I_PT1, It, mean_0, mean_1, unprocess_function):
        It = tf.reshape(It, [-1, 96, 128, 1])
        I_PT1 = tf.reshape(I_PT1, [-1, 96, 128, 1])
        I_PT1 = tf.cast(I_PT1, dtype=tf.float16)

        It = tf.cast(unprocess_function(It, mean_1), dtype=tf.uint16)
        I_PT1 = tf.cast(unprocess_function(I_PT1, mean_0), dtype=tf.uint16)

        result = tf.image.ssim(I_PT1, It, max_val=tf.reduce_max(It) - tf.reduce_min(It))
        mean = tf.reduce_mean(result)

        return mean

    def mask_ssim(self, I_PT1, It, Mt, mean_0, mean_1, unprocess_function):
        It = tf.reshape(It, [-1, 96, 128, 1])
        Mt = tf.reshape(Mt, [-1, 96, 128, 1])
        I_PT1 = tf.reshape(I_PT1, [-1, 96, 128, 1])
        I_PT1 = tf.cast(I_PT1, dtype=tf.float16)

        It = tf.cast(unprocess_function(It, mean_1), dtype=tf.uint16)
        I_PT1 = tf.cast(unprocess_function(I_PT1, mean_0), dtype=tf.uint16)
        Mt = tf.cast(Mt, dtype=tf.uint16)

        mask_image_raw_1 = Mt * It
        mask_output_G1 = Mt * I_PT1
        result = tf.image.ssim(mask_image_raw_1, mask_output_G1, max_val=tf.reduce_max(It) - tf.reduce_min(It))
        mean = tf.reduce_mean(result)
        mean = tf.cast(mean, dtype=tf.float32)

        return mean