-
Notifications
You must be signed in to change notification settings - Fork 3
/
split.py
190 lines (152 loc) · 6.68 KB
/
split.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
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
import tensorflow as tf
import tensorflow.keras.backend as K
from tensorflow.keras.layers import Input,Conv2D, Conv2DTranspose, Dense, Reshape, \
BatchNormalization, GlobalAveragePooling2D, Flatten, Activation
from tensorflow.keras import Model
from tensorflow.keras.optimizers import Adam
import numpy as np
import matplotlib.pyplot as plt
import wget
import os
#dowload pretrained weights
if not(os.path.exists("weights/celeba_split_one.hdf5")):
import wget
url = 'https://www.cs.unibo.it/~asperti/VAEcheckpoint/split_64_4SB_lat150_base128_512_k3.hdf5'
filename = wget.download(url)
os.rename(filename, "weights/celeba_split_one.hdf5")
url = 'https://www.cs.unibo.it/~asperti/VAEcheckpoint/celeba_split_two.hdf5'
filename = wget.download(url)
os.rename(filename, "weights/celeba_split_two.hdf5")
url = 'https://www.cs.unibo.it/~asperti/VAEcheckpoint/celeba_split_three.hdf5'
filename = wget.download(url)
os.rename(filename, "weights/celeba_split_three.hdf5")
def ResBlock(out_dim, depth=2, kernel_size=3, name='ResBlock'):
def body(inputs, **kwargs):
with K.name_scope(name):
y = inputs
for i in range(depth):
y = BatchNormalization(momentum=.999,epsilon=1e-5)(y)
y = Activation('swish')(y) #ReLU()(y)
y = Conv2D(out_dim,kernel_size,padding='same')(y)
s = Conv2D(out_dim, kernel_size,padding='same')(inputs)
return y + s
return(body)
def ResFcBlock(out_dim, depth=2, name='ResFcBlock'):
def body(inputs, **kwargs):
with K.name_scope(name):
y = inputs
for i in range(depth):
y = BatchNormalization(momentum=.999,epsilon=1e-5)(y)
y = Activation('swish')(y) #ReLU()(y)
y = Dense(out_dim)(y)
s = Dense(out_dim)(inputs)
return y + s
return(body)
def ScaleBlock(out_dim, block_per_scale=1, depth_per_block=2, kernel_size=3, name='ScaleBlock'):
def body(inputs, **kwargs):
with K.name_scope(name):
y = inputs
for i in range(block_per_scale):
y = ResBlock(out_dim,depth=depth_per_block, kernel_size=kernel_size)(y)
return y
return (body)
def ScaleFcBlock(out_dim, block_per_scale=1, depth_per_block=2, name='ScaleFcBlock'):
def body(inputs, **kwargs):
with K.name_scope(name):
y = inputs
for i in range(block_per_scale):
y = ResFcBlock(out_dim, depth=depth_per_block)(y)
return y
return(body)
# Model
def Encoder(input_shape, base_dim, kernel_size, num_scale, block_per_scale, depth_per_block,
embedding_dim, name='Encoder'):
with K.name_scope(name):
dim = base_dim
enc_input = Input(shape=input_shape)
y = Conv2D(dim,kernel_size,padding='same',strides=2)(enc_input)
for i in range(num_scale-1):
y = ScaleBlock(dim, block_per_scale, depth_per_block, kernel_size)(y)
if i != num_scale - 1:
dim *= 2
y = Conv2D(dim,kernel_size,strides=2,padding='same')(y)
y = GlobalAveragePooling2D()(y)
ySB = ScaleFcBlock(embedding_dim,1,depth_per_block)(y)
encoder = Model(enc_input,ySB)
return encoder
def Latent(embedding_dim,latent_dim, name="Latent"):
with K.name_scope(name):
emb = Input(shape=embedding_dim)
mu_z = Dense(latent_dim)(emb)
logvar_z = Dense(latent_dim)(emb)
z = mu_z + K.random_normal(shape=K.shape(mu_z)) * K.exp(logvar_z*.5)
back_to_emb = Dense(embedding_dim)
emb_hat = back_to_emb(z)
noise = Input(shape=latent_dim)
emb_gen = back_to_emb(noise)
through_latent = Model(emb,[emb_hat,z,mu_z,logvar_z])
emb_generator = Model(noise,emb_gen)
return through_latent,emb_generator
def Decoder(out_ch, embedding_dim, dims, scales, kernel_size, block_per_scale, depth_per_block, name='Decoder'):
base_wh = 4
data_depth = out_ch
print("dims[0] is = ",dims[0])
print("embedding_dim is ",embedding_dim)
with K.name_scope(name):
emb = Input(shape=(embedding_dim,))
y = Dense(base_wh * base_wh * dims[0])(emb)
y = Reshape((base_wh,base_wh,dims[0]))(y)
for i in range(len(scales) - 1):
y = Conv2DTranspose(dims[i+1], kernel_size, strides=2, padding='same')(y)
y = ScaleBlock(dims[i+1],block_per_scale, depth_per_block, kernel_size)(y)
x_hat = Conv2D(data_depth, kernel_size, 1, padding='same', activation='sigmoid')(y)
decoder = Model(emb,x_hat)
return(decoder)
def FullModel(input_shape,latent_dim,base_dim=32,emb_dim=512, kernel_size=3,num_scale=3,block_per_scale=1,depth_per_block=2):
desired_scale = input_shape[1]
scales, dims = [], []
current_scale, current_dim = 4, base_dim
while current_scale <= desired_scale:
scales.append(current_scale)
dims.append(current_dim)
current_scale *= 2
current_dim = min(current_dim * 2, 1024)
assert (scales[-1] == desired_scale)
dims = list(reversed(dims))
print(dims,scales)
encoder = Encoder(input_shape, base_dim, kernel_size, num_scale, block_per_scale, depth_per_block, emb_dim)
through_latent,emb_generator = Latent(emb_dim,latent_dim)
decoder = Decoder(input_shape[2]*2+1, emb_dim, dims, scales, kernel_size, block_per_scale, depth_per_block)
x = Input(shape=input_shape)
channels = input_shape[2]
gamma = Input(shape=())
emb = encoder(x)
emb_hat, z, z_mean, z_log_var = through_latent(emb)
dec = decoder(emb_hat)
mask = dec[:,:,:,0:1]
img1 = dec[:,:,:,1:1+channels]
img2 = dec[:,:,:,1+channels:1+2*channels]
x_hat = img1*mask + img2*(1-mask)
#x_hat = dec
vae = Model([x,gamma],x_hat)
#loss
beta = 3
L_rec =.5 * K.sum(K.square(x-x_hat), axis=[1,2,3]) / gamma
L_KL = .5 * K.sum(K.square(z_mean) + K.exp(z_log_var) - 1 - z_log_var, axis=-1)
L_tot = K.mean(L_rec + beta * L_KL)
vae.add_loss(L_tot)
return(vae,encoder,decoder,through_latent,emb_generator)
####################################################################
# create model
####################################################################
def get_model(dataset):
if dataset == 'celeba':
latent_dim = 150
input_dim = (64,64,3)
elif dataset == 'cifar10':
latent_dim = 200
input_dim = (32,32,3)
elif dataset == 'mnist':
latent_dim = 32
input_dim = (32,32,1)
return FullModel(input_dim,latent_dim,base_dim=64,num_scale=4,emb_dim=512)