import numpy as np
import os
import math
import tensorflow as tf
import matplotlib.pyplot as plt
from datetime import datetime
from architectures.utils.NN_building_blocks import *
from architectures.utils.NN_gen_building_blocks import *
def lrelu(x, alpha=0.2):
return tf.maximum(alpha*x,x)
# some dummy constants
LEARNING_RATE = None
BETA1 = None
COST_TYPE=None
BATCH_SIZE = None
EPOCHS = None
SAVE_SAMPLE_PERIOD = None
PATH = None
SEED = None
rnd_seed=1
preprocess=None
LAMBDA=.01
EPS=1e-10
CYCL_WEIGHT=None
GAN_WEIGHT=None
DISCR_STEPS=None
GEN_STEPS=None
max_true=None
max_reco=None
n_H_A=None
n_W_A=None
n_W_B=None
n_H_B=None
n_C=None
d_sizes_A=None
d_sizes_B=None
g_sizes_A=None
g_sizes_B=None
class cycleGAN_fullresidual(object):
def __init__(
self,
n_H_A=n_H_A, n_W_A=n_W_A,
n_H_B=n_H_B, n_W_B=n_W_B, n_C=n_C,
max_true=max_true, max_reco=max_reco,
d_sizes_A=d_sizes_A, d_sizes_B=d_sizes_B, g_sizes_A=g_sizes_A, g_sizes_B=g_sizes_B,
lr=LEARNING_RATE, beta1=BETA1, preprocess=preprocess,
cost_type=COST_TYPE, gan_weight=GAN_WEIGHT, cycl_weight=CYCL_WEIGHT,
discr_steps=DISCR_STEPS, gen_steps=GEN_STEPS,
batch_size=BATCH_SIZE, epochs=EPOCHS,
save_sample=SAVE_SAMPLE_PERIOD, path=PATH, seed=SEED,
):
"""
Positional arguments:
- width of (square) image
- number of channels of input image
- discriminator sizes
a python dict of the kind
d_sizes = { 'convblocklayer_n':[(n_c+1, kernel, stride, apply_batch_norm, weight initializer),
(,,,,),
(,,,,),
],
'convblock_shortcut_layer_n':[(,,,)],
'dense_layers':[(n_o, apply_bn, weight_init)]
}
- generator sizes
a python dictionary of the kind
g_sizes = {
'z':latent_space_dim,
'projection': int,
'bn_after_project':bool
'deconvblocklayer_n':[(n_c+1, kernel, stride, apply_batch_norm, weight initializer),
(,,,,),
(,,,,),
],
'deconvblock_shortcut_layer_n':[(,,,)],
'dense_layers':[(n_o, apply_bn, weight_init)]
'activation':function
}
Keyword arguments:
- lr = LEARNING_RATE (float32)
- beta1 = ema parameter for adam opt (float32)
- batch_size (int)
- save_sample = after how many batches iterations to save a sample (int)
- path = relative path for saving samples
"""
self.max_true=max_true
self.max_reco=max_reco
self.seed=seed
self.n_W_A = n_W_A
self.n_H_A = n_H_A
self.n_W_B = n_W_B
self.n_H_B = n_H_B
self.n_C = n_C
#input data
self.input_A = tf.placeholder(
tf.float32,
shape=(None,
n_H_A, n_W_A, n_C),
name='X_A',
)
self.input_B = tf.placeholder(
tf.float32,
shape=(None,
n_H_B, n_W_B, n_C),
name='X_B',
)
self.batch_sz = tf.placeholder(
tf.int32,
shape=(),
name='batch_sz'
)
self.lr = tf.placeholder(
tf.float32,
shape=(),
name='lr'
)
D_A = Discriminator(self.input_A, d_sizes_A, 'A')
D_B = Discriminator(self.input_B, d_sizes_B, 'B')
G_A_to_B = cycleGenerator_fullresidual(self.input_A, self.n_H_B, self.n_W_B, g_sizes_A, 'A_to_B')
G_B_to_A = cycleGenerator_fullresidual(self.input_B, self.n_H_A, self.n_W_A, g_sizes_B, 'B_to_A')
#first cycle (A to B)
with tf.variable_scope('discriminator_A') as scope:
logits_A = D_A.d_forward(self.input_A)
with tf.variable_scope('generator_A_to_B') as scope:
sample_images_B = G_A_to_B.g_forward(self.input_A)
#second cycle (B to A)
with tf.variable_scope('discriminator_B') as scope:
logits_B = D_B.d_forward(self.input_B)
with tf.variable_scope('generator_B_to_A') as scope:
sample_images_A = G_B_to_A.g_forward(self.input_B)
with tf.variable_scope('discriminator_A') as scope:
scope.reuse_variables()
sample_logits_A = D_A.d_forward(sample_images_A, reuse=True)
with tf.variable_scope('discriminator_B') as scope:
scope.reuse_variables()
sample_logits_B = D_B.d_forward(sample_images_B, reuse=True)
with tf.variable_scope('generator_A_to_B') as scope:
scope.reuse_variables()
cycl_B = G_A_to_B.g_forward(sample_images_A, reuse=True)
with tf.variable_scope('generator_B_to_A') as scope:
scope.reuse_variables()
cycl_A = G_B_to_A.g_forward(sample_images_B, reuse=True)
self.input_test_A = tf.placeholder(
tf.float32,
shape=(None,
n_H_A, n_W_A, n_C),
name='X_test_A',
)
# get sample images for test time
with tf.variable_scope('generator_A_to_B') as scope:
scope.reuse_variables()
self.sample_images_test_A_to_B = G_A_to_B.g_forward(
self.input_test_A, reuse=True, is_training=False
)
self.input_test_B = tf.placeholder(
tf.float32,
shape=(None,
n_H_B, n_W_B, n_C),
name='X_test_B',
)
with tf.variable_scope('generator_B_to_A') as scope:
scope.reuse_variables()
self.sample_images_test_B_to_A = G_B_to_A.g_forward(
self.input_test_B, reuse=True, is_training=False
)
#parameters lists
self.d_params_A =[t for t in tf.trainable_variables() if 'discriminator_A' in t.name]
self.d_params_B =[t for t in tf.trainable_variables() if 'discriminator_B' in t.name]
self.g_params_A =[t for t in tf.trainable_variables() if 'B_to_A' in t.name]
self.g_params_B =[t for t in tf.trainable_variables() if 'A_to_B' in t.name]
#cost building
if cost_type == 'GAN':
#Discriminators cost
#Discriminator_A cost
#cost is low if real images are predicted as real (1)
d_cost_real_A = tf.nn.sigmoid_cross_entropy_with_logits(
logits=logits_A,
labels=tf.ones_like(logits_A)
)
# #cost is low if fake generated images are predicted as fake (0)
d_cost_fake_A = tf.nn.sigmoid_cross_entropy_with_logits(
logits=sample_logits_A,
labels=tf.zeros_like(sample_logits_A)
)
self.d_cost_A = tf.reduce_mean(d_cost_real_A) + tf.reduce_mean(d_cost_fake_A)
#Discriminator_B cost
d_cost_real_B = tf.nn.sigmoid_cross_entropy_with_logits(
logits=logits_B,
labels=tf.ones_like(logits_B)
)
d_cost_fake_B = tf.nn.sigmoid_cross_entropy_with_logits(
logits=sample_logits_B,
labels=tf.zeros_like(sample_logits_B)
)
self.d_cost_B = tf.reduce_mean(d_cost_real_B) + tf.reduce_mean(d_cost_fake_B)
#Generator cost
#cost is low if logits from discriminator A on samples generated by G_B_to_A
#are predicted as true (1)
self.g_GAN_cost_A = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(
logits=sample_logits_A,
labels=tf.ones_like(sample_logits_A)
)
)
#cost is low if logits from discriminator B on samples generated by G_A_to_B
#are predicted as true (1)
self.g_GAN_cost_B = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(
logits=sample_logits_B,
labels=tf.ones_like(sample_logits_B)
)
)
#cycle cost is low if cyclic images are similar to input images (in both sets)
self.g_cycle_cost_A = tf.reduce_mean(tf.abs(self.input_A-cycl_A))
self.g_cycle_cost_B = tf.reduce_mean(tf.abs(self.input_B-cycl_B))
g_cycle_cost= self.g_cycle_cost_A+self.g_cycle_cost_B
self.g_cost_A = gan_weight*self.g_GAN_cost_A + cycl_weight*g_cycle_cost
self.g_cost_B = gan_weight*self.g_GAN_cost_B + cycl_weight*g_cycle_cost
if cost_type == 'WGAN-gp':
#Discriminators cost
#Discriminator A
self.d_cost_A = tf.reduce_mean(sample_logits_A) - tf.reduce_mean(logits_A)
self.d_cost_B = tf.reduce_mean(sample_logits_B) - tf.reduce_mean(logits_B)
self.g_cost_A = -tf.reduce_mean(sample_logits_A)
self.g_cost_B = -tf.reduce_mean(sample_logits_B)
# g_cycle_cost_A = tf.reduce_mean(tf.abs(self.input_A-cycl_A))
# g_cycle_cost_B = tf.reduce_mean(tf.abs(self.input_B-cycl_B))
# g_cycle_cost= g_cycle_cost_A+g_cycle_cost_B
# self.g_cost_A = g_cost_A + 0*g_cycle_cost
# self.g_cost_B = g_cost_B + 0*g_cycle_cost
alpha_A = tf.random_uniform(
shape=[self.batch_sz,self.n_H_A,self.n_W_A,self.n_C],
minval=0.,
maxval=1.
)
interpolates_A = alpha_A*self.input_A+(1-alpha_A)*sample_images_A
with tf.variable_scope('discriminator_A') as scope:
scope.reuse_variables()
disc_A_interpolates = D_A.d_forward(interpolates_A,reuse = True)
gradients_A = tf.gradients(disc_A_interpolates,[interpolates_A])[0]
slopes_A = tf.sqrt(tf.reduce_sum(tf.square(gradients_A), reduction_indices=[1]))
gradient_penalty_A = tf.reduce_mean((slopes_A-1)**2)
self.d_cost_A+=LAMBDA*gradient_penalty_A
alpha_B = tf.random_uniform(
shape=[self.batch_sz,self.n_H_B,self.n_W_B,self.n_C],
minval=0.,
maxval=1.
)
interpolates_B = alpha_B*self.input_B+(1-alpha_B)*sample_images_B
with tf.variable_scope('discriminator_B') as scope:
scope.reuse_variables()
disc_B_interpolates = D_B.d_forward(interpolates_B,reuse = True)
gradients_B = tf.gradients(disc_B_interpolates,[interpolates_B])[0]
slopes_B = tf.sqrt(tf.reduce_sum(tf.square(gradients_B), reduction_indices=[1]))
gradient_penalty_B = tf.reduce_mean((slopes_B-1)**2)
self.d_cost_B+=LAMBDA*gradient_penalty_B
self.d_train_op_A = tf.train.AdamOptimizer(
learning_rate=lr,
beta1=beta1,
beta2=0.9,
).minimize(
self.d_cost_A,
var_list=self.d_params_A
)
self.d_train_op_B = tf.train.AdamOptimizer(
learning_rate=lr,
beta1=beta1,
beta2=0.9,
).minimize(
self.d_cost_B,
var_list=self.d_params_B
)
self.g_train_op_A = tf.train.AdamOptimizer(
learning_rate=lr,
beta1=beta1,
beta2=0.9,
).minimize(
self.g_cost_A,
var_list=self.g_params_A
)
self.g_train_op_B = tf.train.AdamOptimizer(
learning_rate=lr,
beta1=beta1,
beta2=0.9,
).minimize(
self.g_cost_B,
var_list=self.g_params_B
)
#Measure accuracy of the discriminators
real_predictions_A = tf.cast(logits_A>0,tf.float32)
fake_predictions_A = tf.cast(sample_logits_A<0,tf.float32)
num_predictions=2.0*batch_size
num_correct_A = tf.reduce_sum(real_predictions_A)+tf.reduce_sum(fake_predictions_A)
self.d_accuracy_A = num_correct_A/num_predictions
real_predictions_B = tf.cast(logits_B>0,tf.float32)
fake_predictions_B = tf.cast(sample_logits_B<0,tf.float32)
num_predictions=2.0*batch_size
num_correct_B = tf.reduce_sum(real_predictions_B)+tf.reduce_sum(fake_predictions_B)
self.d_accuracy_B = num_correct_B/num_predictions
self.batch_size=batch_size
self.epochs=epochs
self.save_sample=save_sample
self.path=path
self.lr = lr
self.D_A=D_A
self.D_B=D_B
self.G_A_to_B=G_A_to_B
self.G_B_to_A=G_B_to_A
self.sample_images_B=sample_images_B
self.sample_images_A=sample_images_A
self.preprocess=preprocess
self.cost_type=cost_type
self.gen_steps=gen_steps
self.discr_steps=discr_steps
self.cycl_weight=cycl_weight
def set_session(self, session):
self.session = session
for block in self.D_A.d_conv_layers:
block.set_session(session)
for layer in self.D_A.d_dense_layers:
layer.set_session(session)
for layer in self.G_A_to_B.g_blocks:
layer.set_session(session)
for block in self.D_B.d_conv_layers:
block.set_session(session)
for layer in self.D_B.d_dense_layers:
layer.set_session(session)
for block in self.G_B_to_A.g_blocks:
block.set_session(session)
def fit(self, X_A, X_B, validating_size):
all_A = X_A
all_B = X_B
gen_steps = self.gen_steps
discr_steps = self.discr_steps
m = X_A.shape[0]
train_A = all_A[0:m-validating_size]
train_B = all_B[0:m-validating_size]
validating_A = all_A[m-validating_size:m]
validating_B = all_B[m-validating_size:m]
seed = self.seed
d_costs_A=[]
d_gps_A=[]
g_GANs_A=[]
g_cycles_A=[]
g_costs_A=[]
d_costs_B=[]
d_gps_B=[]
g_GANs_B=[]
g_cycles_B=[]
g_costs_B=[]
N = len(train_A)
n_batches = N // self.batch_size
total_iters=0
print('\n ****** \n')
print('Training cycle GAN with residual arichitecture and a total of ' +str(N)+' samples distributed in '+ str((N)//self.batch_size) +' batches of size '+str(self.batch_size)+'\n')
print('The validation set consists of {0} images'.format(validating_A.shape[0]))
print('The learning rate set is '+str(self.lr)+', and every ' +str(self.save_sample)+ ' batches a generated sample will be saved to '+ self.path)
print('\n ****** \n')
for epoch in range(self.epochs):
seed +=1
print('Epoch:', epoch)
batches_A = unsupervised_random_mini_batches(X_A, self.batch_size, seed)
batches_B = unsupervised_random_mini_batches(X_B, self.batch_size, seed)
for X_batch_A, X_batch_B in zip(batches_A,batches_B):
bs = X_batch_A.shape[0]
t0 = datetime.now()
#optimize generator_A
g_cost_A=0
g_GAN_A=0
g_cycle_A=0
for i in range(gen_steps):
_, g_cost_A, g_GAN_A, g_cycle_A = self.session.run(
(self.g_train_op_A, self.g_cost_A, self.g_GAN_cost_A, self.g_cycle_cost_A),
feed_dict={self.input_A:X_batch_A, self.input_B:X_batch_B, self.batch_sz:bs},
)
g_cost_A+=g_cost_A
g_GAN_A+=g_GAN_A
g_cycle_A+=g_cycle_A
g_costs_A.append(g_cost_A/gen_steps)
g_GANs_A.append(g_GAN_A/gen_steps)
g_cycles_A.append(self.cycl_weight*g_cycle_A/gen_steps)
#optimize discriminator_B
d_cost_B=0
d_gp_B=0
for i in range(discr_steps):
if self.cost_type=='WGAN-gp':
_, d_cost_B, d_acc_B, d_gp_B = self.session.run(
(self.d_train_op_B, self.d_cost_B, self.d_accuracy_B, self.gradient_penalty_B),
feed_dict={self.input_A:X_batch_A, self.input_B:X_batch_B, self.batch_sz:bs},
)
d_gp_B+=d_gp_B
else:
_, d_cost_B, d_acc_B, = self.session.run(
(self.d_train_op_B, self.d_cost_B, self.d_accuracy_B),
feed_dict={self.input_A:X_batch_A, self.input_B:X_batch_B, self.batch_sz:bs},
)
d_cost_B+=d_cost_B
d_costs_B.append(d_cost_B/discr_steps)
d_gps_B.append(LAMBDA*d_gp_B/discr_steps)
#optimize generator_B
g_cost_B=0
g_GAN_B=0
g_cycle_B=0
for i in range(gen_steps):
_, g_cost_B, g_GAN_B, g_cycle_B = self.session.run(
(self.g_train_op_B, self.g_cost_B, self.g_GAN_cost_B, self.g_cycle_cost_B),
feed_dict={self.input_A:X_batch_A, self.input_B:X_batch_B, self.batch_sz:bs},
)
g_cost_B+=g_cost_B
g_GAN_B+=g_GAN_B
g_cycle_B+=g_cycle_B
g_costs_B.append(g_cost_B/gen_steps)
g_GANs_B.append(g_GAN_B/gen_steps)
g_cycles_B.append(self.cycl_weight*g_cycle_B/gen_steps)
d_cost_A=0
d_gp_A=0
for i in range(discr_steps):
#optimize Discriminator_A
if self.cost_type=='WGAN-gp':
_, d_cost_A, d_acc_A, d_gp_A = self.session.run(
(self.d_train_op_A, self.d_cost_A, self.d_accuracy_A, self.gradient_penalty_A),
feed_dict={self.input_A:X_batch_A, self.input_B:X_batch_B, self.batch_sz:bs},
)
d_gp_A+=d_gp_A
else:
_, d_cost_A, d_acc_A, = self.session.run(
(self.d_train_op_A, self.d_cost_A, self.d_accuracy_A),
feed_dict={self.input_A:X_batch_A, self.input_B:X_batch_B, self.batch_sz:bs},
)
d_cost_A+=d_cost_A
d_costs_A.append(d_cost_A/discr_steps)
d_gps_A.append(LAMBDA*d_gp_A/discr_steps)
total_iters += 1
if total_iters % self.save_sample ==0:
print("At iter: %d - dt: %s - d_acc_A: %.2f" % (total_iters, datetime.now() - t0, d_acc_A))
print("At iter: %d - dt: %s - d_acc_B: %.2f" % (total_iters, datetime.now() - t0, d_acc_B))
print("Discrimator_A cost {0:.4g}, Generator_A_to_B cost {1:.4g}".format(d_cost_A, g_cost_A))
print("Discrimator_B cost {0:.4g}, Generator_B_to_A cost {1:.4g}".format(d_cost_B, g_cost_B))
print('Saving a sample...')
#A is true
#B is reco
if self.preprocess=='normalise':
draw_nn_sample(validating_A, validating_B, 1, self.preprocess,
self.max_true, self.max_reco,
f=self.get_sample_A_to_B, is_training=True,
total_iters=total_iters, PATH=self.path)
else:
draw_nn_sample(validating_A, validating_B, 1, self.preprocess,
f=self.get_sample_A_to_B, is_training=True,
total_iters=total_iters, PATH=self.path)
plt.clf()
plt.subplot(1,2,1)
plt.plot(d_costs_A, label='Discriminator A cost')
plt.plot(d_gps_A, label='Gradient penalty A')
plt.plot(g_GANs_A, label='Generator B to A GAN cost')
plt.xlabel('Epoch')
plt.ylabel('Cost')
plt.legend()
plt.subplot(1,2,2)
plt.plot(g_costs_A, label='Generator B to A total cost')
plt.plot(g_GANs_A, label='Generator B to A GAN cost')
plt.plot(g_cycles_B, label='Generators B to A to B cycle cost')
plt.xlabel('Epoch')
plt.ylabel('Cost')
plt.legend()
fig = plt.gcf()
fig.set_size_inches(15,5)
plt.savefig(self.path+'/cost_iteration_gen_disc_B_to_A.png',dpi=150)
plt.clf()
plt.subplot(1,2,1)
plt.plot(d_costs_B, label='Discriminator B cost')
plt.plot(d_gps_B, label='Gradient penalty B')
plt.plot(g_GANs_B, label='Generator A to B GAN cost')
plt.xlabel('Epoch')
plt.ylabel('Cost')
plt.legend()
plt.subplot(1,2,2)
plt.plot(g_costs_B, label='Generator A to B total cost')
plt.plot(g_GANs_B, label='Generator A to B GAN cost')
plt.plot(g_cycles_B, label='Generators B to A to B cycle cost')
plt.xlabel('Epoch')
plt.ylabel('Cost')
plt.legend()
fig = plt.gcf()
fig.set_size_inches(15,5)
plt.savefig(self.path+'/cost_iteration_gen_disc_A_to_B.png',dpi=150)
def get_sample_A_to_B(self, Z):
one_sample = self.session.run(
self.sample_images_test_A_to_B,
feed_dict={self.input_test_A:Z, self.batch_sz: 1})
return one_sample
def get_sample_B_to_A(self, Z):
one_sample = self.session.run(
self.sample_images_test_B_to_A,
feed_dict={self.input_test_B:Z, self.batch_sz: 1})
return one_sample
def get_samples_A_to_B(self, Z):
many_samples = self.session.run(
self.sample_images_test_A_to_B,
feed_dict={self.input_test_A:Z, self.batch_sz: Z.shape[0]})
return many_samples
def get_samples_B_to_A(self, Z):
many_samples = self.session.run(
self.sample_images_test_B_to_A,
feed_dict={self.input_test_B:Z, self.batch_sz: Z.shape[0]})
return many_samples
Davide Lancierini