Python get_image_batch Examples

Example #1

A_DIR = '/hdd/cs599/spectro/testA/*'
B_DIR = '/hdd/cs599/spectro/testB/*'
RESULT_A_DIR = '/hdd/cs599/spectro/results/resultA/'
RESULT_B_DIR = '/hdd/cs599/spectro/results/resultB/'
if not os.path.exists(RESULT_A_DIR): os.makedirs(RESULT_A_DIR)
if not os.path.exists(RESULT_B_DIR): os.makedirs(RESULT_B_DIR)

# Remove files in the output
for removeFile in chain(glob.glob(RESULT_A_DIR + '*'),
                        glob.glob(RESULT_B_DIR + '*')):
    os.remove(removeFile)

#############################################3
# Define Network
#############################################
f_a, a = dataset.get_image_batch(A_DIR, BATCH_SIZE, train=False)
f_b, b = dataset.get_image_batch(B_DIR, BATCH_SIZE, train=False)

with tf.variable_scope('gen_a_to_b') as a_to_b_scope:
    b_gen = build_enc_dec(a)
with tf.variable_scope('gen_b_to_a') as b_to_a_scope:
    a_gen = build_enc_dec(b)

with tf.variable_scope('gen_b_to_a', reuse=True):
    a_identity = build_enc_dec(b_gen, True)
with tf.variable_scope('gen_a_to_b', reuse=True):
    b_identity = build_enc_dec(a_gen, True)

#################################
# Miscellaneous(summary, init, etc.)
#################################

Example #2

LAMBDA = 10
LAMBDA_CYCLE = 10

BATCH_SIZE = 8

MAX_ITERATION = 1000000
SAVE_PERIOD = 10000
SUMMARY_PERIOD = 50

NUM_CRITIC_TRAIN = 4

#############################################3
# Define Network
#############################################3
_, a = dataset.get_image_batch(A_DIR, BATCH_SIZE)
_, b = dataset.get_image_batch(B_DIR, BATCH_SIZE)

with tf.variable_scope('gen_a_to_b') as a_to_b_scope:
    b_gen = build_enc_dec(a)
with tf.variable_scope('gen_b_to_a') as b_to_a_scope:
    a_gen = build_enc_dec(b)

with tf.variable_scope('gen_b_to_a', reuse=True):
    a_identity = build_enc_dec(b_gen, True)
with tf.variable_scope('gen_a_to_b', reuse=True):
    b_identity = build_enc_dec(a_gen, True)

with tf.variable_scope('c_a') as scope:
    alpha = tf.random_uniform(shape=[BATCH_SIZE, 1, 1, 1],
                              minval=0.,

Example #3

File: inference.py Project: QianQQ/Voice-Conversion

BATCH_SIZE = 1
A_DIR = '/hdd/cs599/spectro/testA/*'
B_DIR = '/hdd/cs599/spectro/testB/*'
RESULT_A_DIR = '/hdd/cs599/spectro/results/resultA/'
RESULT_B_DIR = '/hdd/cs599/spectro/results/resultB/'
if not os.path.exists(RESULT_A_DIR): os.makedirs(RESULT_A_DIR)
if not os.path.exists(RESULT_B_DIR): os.makedirs(RESULT_B_DIR)

# Remove files in the output
for removeFile in chain(glob.glob(RESULT_A_DIR + '*'), glob.glob(RESULT_B_DIR + '*')):
    os.remove(removeFile)

#############################################3
# Define Network
#############################################
f_a, a = dataset.get_image_batch(A_DIR, BATCH_SIZE, train=False)
f_b, b = dataset.get_image_batch(B_DIR, BATCH_SIZE, train=False)

with tf.variable_scope('gen_a_to_b') as a_to_b_scope :
    b_gen = build_enc_dec(a)
with tf.variable_scope('gen_b_to_a') as b_to_a_scope :
    a_gen = build_enc_dec(b)

with tf.variable_scope('gen_b_to_a',reuse=True) :
    a_identity = build_enc_dec(b_gen,True)
with tf.variable_scope('gen_a_to_b',reuse=True) :
    b_identity = build_enc_dec(a_gen,True)

#################################
# Miscellaneous(summary, init, etc.)
#################################

Example #4

File: main.py Project: QianQQ/Voice-Conversion

LAMBDA = 10
LAMBDA_CYCLE = 10

BATCH_SIZE = 8

MAX_ITERATION = 1000000
SAVE_PERIOD = 10000
SUMMARY_PERIOD = 50

NUM_CRITIC_TRAIN = 4

#############################################3
# Define Network
#############################################3
_, a = dataset.get_image_batch(A_DIR, BATCH_SIZE)
_, b = dataset.get_image_batch(B_DIR, BATCH_SIZE)

with tf.variable_scope('gen_a_to_b') as a_to_b_scope :
    b_gen = build_enc_dec(a)
with tf.variable_scope('gen_b_to_a') as b_to_a_scope :
    a_gen = build_enc_dec(b)

with tf.variable_scope('gen_b_to_a',reuse=True) :
    a_identity = build_enc_dec(b_gen,True)
with tf.variable_scope('gen_a_to_b',reuse=True) :
    b_identity = build_enc_dec(a_gen,True)

with tf.variable_scope('c_a') as scope:
    alpha = tf.random_uniform(shape=[BATCH_SIZE,1,1,1], minval=0.,maxval=1.)
    a_hat = alpha * a + (1.0 - alpha) * a_gen