-
Notifications
You must be signed in to change notification settings - Fork 10
/
3dgan_train.py
244 lines (187 loc) · 7.96 KB
/
3dgan_train.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
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
# -*- coding: utf-8 -*-
import cv2
from numpy import zeros
from numpy import ones
from numpy import expand_dims
from numpy.random import randn
from numpy.random import randint
from keras.datasets.fashion_mnist import load_data
from keras.optimizers import Adam
from keras.models import Model,Sequential
from keras.layers import Input
from keras.layers import Dense
from keras.layers import Reshape
from keras.layers import Flatten
from keras.layers import Conv2D,Conv1D,multiply
from keras.layers import Conv2DTranspose
from keras.layers import LeakyReLU
from keras.layers import Dropout
from keras.layers import Embedding
from keras.layers import Activation
from keras.layers import Concatenate
from keras.layers.convolutional import UpSampling2D, Conv2D, UpSampling1D
from keras.initializers import RandomNormal
from matplotlib import pyplot
from sklearn.decomposition import PCA
from sklearn.preprocessing import StandardScaler
import pandas as pd
import seaborn as sns
from skimage.color import label2rgb
import matplotlib.pyplot as plt
import keras.backend as K
K.set_learning_phase(1)
from keras import applications, layers
from numpy.random import seed
from tensorflow import set_random_seed
seed(42)
set_random_seed(42)
#load dataset
import numpy as np
data = np.load('dataset/ds.npy')
gt = np.load('dataset/gt.npy')
H = data.shape[0]
W = data.shape[1]
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from patchify import patchify, unpatchify
def define_discriminator(in_shape=(64,64,3), n_classes=17):
model = Sequential()
model.add(Conv2D(64, kernel_size=(4,4), strides=2, input_shape=in_shape, padding="same"))
model.add(LeakyReLU(alpha=0.2))
model.add(Conv2D(128, kernel_size=(4,4), strides=2, padding="same"))
model.add(LeakyReLU(alpha=0.2))
model.add(BatchNormalization(momentum=0.8))
model.add(Conv2D(256, kernel_size=(4,4), strides=2, padding="same"))
model.add(LeakyReLU(alpha=0.2))
model.add(BatchNormalization(momentum=0.8))
model.add(Conv2D(512, kernel_size=(4,4), strides=2, padding="same"))
model.add(LeakyReLU(alpha=0.2))
model.add(BatchNormalization(momentum=0.8))
model.add(Conv2D(64, kernel_size=(4,4), strides=1, padding="same"))
model.add(Flatten())
img = Input(shape=in_shape)
features = model(img)
validity = Dense(1, activation="sigmoid")(features)
label = Dense(n_classes, activation="softmax")(features)
model = Model(img, [validity, label])
opt = Adam(lr=0.0002, beta_1=0.5)
model.summary()
model.compile(loss=['binary_crossentropy', 'sparse_categorical_crossentropy'], optimizer=opt,metrics=['accuracy'])
return model
def define_generator(latent_dim, n_classes=17):
model = Sequential()
model.add(Dense(2*2*512, activation="relu", input_dim=latent_dim))
model.add(Reshape((2,2, 512)))
model.add(BatchNormalization(momentum=0.8))
# model.add(UpSampling2D(size=4))
model.add(Conv2DTranspose(512, kernel_size=(4,4), strides=2, padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(momentum=0.8))
# model.add(UpSampling2D(size=4))
model.add(Conv2DTranspose(256, kernel_size=(4,4), strides=2, padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(momentum=0.8))
# model.add(UpSampling2D(size=4))
model.add(Conv2DTranspose(128, kernel_size=(4,4), strides=2, padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(momentum=0.8))
# model.add(UpSampling2D(size=4))
model.add(Conv2DTranspose(64, kernel_size=(4,4), strides=2, padding="same"))
model.add(Activation("relu"))
model.add(BatchNormalization(momentum=0.8))
# model.add(UpSampling2D(size=4))
model.add(Conv2DTranspose(3, kernel_size=(4,4), strides=2, padding='same'))
model.add(Activation("tanh"))
model.summary()
noise = Input(shape=(latent_dim,))
label = Input(shape=(1,), dtype='int32')
label_embedding = Flatten()(Embedding(n_classes, latent_dim)(label))
model_input = multiply([noise, label_embedding])
img = model(model_input)
model.summary()
return Model([noise, label], img)
voxel = data.reshape(-1,data.shape[2])
scaler = StandardScaler()
voxel = scaler.fit_transform(voxel)
pca = PCA(n_components=3)
principalComponents = pca.fit_transform(voxel)
principalDf = pd.DataFrame(data = principalComponents, columns = ['pc'+str(x) for x in range(3)])
fullDataX = principalComponents.reshape(H,W,3)
fullDataY = gt
paddedDatax = cv2.copyMakeBorder( fullDataX, 32, 31, 32, 31, cv2.BORDER_REPLICATE)
X_train = patchify(paddedDatax, (64, 64, 3), step=1).reshape(-1,64,64,3)
y_train = fullDataY.reshape(-1,)
X_train, X_test, y_train, y_test = train_test_split(X_train, y_train, test_size=0.50)
def generate_real_test_samples(n_samples):
images, labels = X_test, y_test
ix = randint(0, images.shape[0], n_samples)
X, labels = images[ix], labels[ix]
y = ones((n_samples, 1))
return [X, labels], y
def generate_real_samples(n_samples):
images, labels = X_train, y_train
ix = randint(0, images.shape[0], n_samples)
X, labels = images[ix], labels[ix]
y = ones((n_samples, 1))
return [X, labels], y
def generate_latent_points(latent_dim, n_samples, n_classes=17):
x_input = randn(latent_dim * n_samples)
z_input = x_input.reshape(n_samples, latent_dim)
labels = randint(0, n_classes, n_samples)
return [z_input, labels]
def generate_fake_samples(generator, latent_dim, n_samples):
z_input, labels_input = generate_latent_points(latent_dim, n_samples)
images = generator.predict([z_input, labels_input])
y = zeros((n_samples, 1))
return [images, labels_input], y
def summarize_performance(step, g_model, latent_dim):
filename = 'model_%04d.h5' % (step+1)
g_model.save_weights(filename)
!cp -f *.h5 /content/drive/My\ Drive/models
def train(g_model, d_model, gan_model, latent_dim, n_epochs=600, n_batch=4096):
import warnings
warnings.filterwarnings("ignore")
bat_per_epo = int(10000 / n_batch)
n_steps = bat_per_epo * n_epochs
half_batch = int(n_batch / 2)
prev_acc1 = 0.20
i = 0
while True:
[X_real, labels_real], y_real = generate_real_samples( half_batch)
d_loss_real = d_model.train_on_batch(X_real, [y_real, labels_real])
[X_fake, labels_fake], y_fake = generate_fake_samples(g_model, latent_dim, half_batch)
d_loss_fake = d_model.train_on_batch(X_fake, [y_fake, labels_fake])
d_loss = 0.5 * np.add(d_loss_real, d_loss_fake)
[z_input, z_labels] = generate_latent_points(latent_dim, half_batch)
y_gan = ones((half_batch, 1))
g_loss = gan_model.train_on_batch([z_input, z_labels], [y_gan, z_labels])
op_acc = d_loss[4]
trainHist.append([d_loss,g_loss])
print ("Training Metrics: %d [D loss: %f, acc.: %.2f%%, op_acc: %.2f%%] [G loss: %f]" % (i, d_loss[0], 100*d_loss[3], 100*d_loss[4], g_loss[0]))
[X_treal, labels_treal], y_treal = generate_real_test_samples(half_batch)
d_loss_real = d_model.test_on_batch(X_treal, [y_treal, labels_treal])
[X_fake, labels_fake], y_fake = generate_fake_samples(g_model, latent_dim, half_batch)
d_loss_fake = d_model.test_on_batch(X_fake, [y_fake, labels_fake])
d_loss = 0.5 * np.add(d_loss_real, d_loss_fake)
[z_input, z_labels] = generate_latent_points(latent_dim, n_batch)
y_gan = ones((n_batch, 1))
g_loss = gan_model.test_on_batch([z_input, z_labels], [y_gan, z_labels])
validationHist.append([d_loss,g_loss])
print ("Validation Metrics: %d [D loss: %f, acc.: %.2f%%, op_acc: %.2f%%] [G loss: %f]" % (i, d_loss[0], 100*d_loss[3], 100*d_loss[4], g_loss[0]))
if (op_acc > 0.95 and d_loss[4] > 0.95) or i > 6000:
prev_acc1 = op_acc
prev_acc2 = d_loss[4]
summarize_performance(i, d_model, latent_dim)
break
if (i+1) % (bat_per_epo * 10) == 0:
summarize_performance(i, d_model, latent_dim)
i += 1
latent_dim = 100
discriminator = define_discriminator()
generator = define_generator(latent_dim)
discriminator.trainable = False
gan_output = discriminator(generator.output)
gan_model = Model(generator.input, gan_output)
opt = Adam(lr=0.002, beta_1=0.5)
gan_model.compile(loss=['binary_crossentropy', 'sparse_categorical_crossentropy'], optimizer=opt)
train(generator, discriminator, gan_model, latent_dim)