CharacterAutoencoder/OmniglotAutoencoderTrain.py at master · AlexGreason/CharacterAutoencoder · 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
import numpy as np
from keras.layers import Input, Dense, Conv2D, MaxPooling2D, Reshape, Flatten, UpSampling2D
from keras.models import Model


nb_epoch = 1000

batch_size = 128
sidelen = 96
original_shape = (None, 1, sidelen, sidelen)
latent_dim = 16
intermediate_dim = 256

x = Input(batch_shape=original_shape)
a = Conv2D(128, (5, 5), padding='same', activation='relu')(x)
b = MaxPooling2D(pool_size=(4, 4))(a)
c = Conv2D(128, (3,3), padding='same', activation='relu')(b)
d = Conv2D(16, (3,3), padding='same', activation='relu')(c)
d_reshaped = Flatten()(d)
h = Dense(intermediate_dim, activation='relu')(d_reshaped)
z_mean = Dense(latent_dim)(h)

# we instantiate these layers separately so as to reuse them later
decoder_h = Dense(intermediate_dim, activation='relu')
i = Dense(8 * 24 * 24, activation='relu')
j = Reshape((8, 24, 24))
k = Conv2D(128, (3,3), padding='same', activation='relu')
l = UpSampling2D((4, 4))
m = Conv2D(128, (3,3), padding='same', activation='relu')
n = Conv2D(128, (3,3), padding='same', activation='relu')
decoder_mean = Conv2D(1, (3,3), padding='same', activation='sigmoid')

h_decoded = decoder_h(z_mean)
i_decoded = i(h_decoded)
j_decoded = j(i_decoded)
k_decoded = k(j_decoded)
l_decoded = l(k_decoded)
m_decoded = m(l_decoded)
n_decoded = n(m_decoded)
x_decoded_mean = decoder_mean(n_decoded)

vae = Model(x, x_decoded_mean)
vae.compile(optimizer='rmsprop', loss='binary_crossentropy')

file = "omniglot_16_"
versionnum = 2


computer = "desktop"

if computer == "laptop":
    x_train = np.load("/home/exa/Documents/PythonData/images_all_processed.npy")

elif computer == "desktop":
    x_train = np.load("/media/exa/Archival drive/conlangstuff/images_all_processed.npy").astype("float16")

x_train = x_train.reshape((x_train.shape[0], 1, sidelen, sidelen))

vae.load_weights("omniglot_16_1.sav")
if nb_epoch > 0:
    for epoch in range(nb_epoch):
        print("epoch", epoch)
        vae.fit(x_train, x_train,
                shuffle=True,
                verbose=1,
                nb_epoch=1,
                batch_size=batch_size,
                validation_split=.1)
        json_string = vae.to_json()
        open(file + str(versionnum) + ".json", 'w').write(json_string)
        vae.save_weights(file + str(versionnum) + ".sav", overwrite=True)