def create(self):
self.X = tf.reshape(
self.X,
shape=[-1, self.IMAGE_SIZE, self.IMAGE_SIZE, self.CHANEELS])
conv1 = util.conv_layer(self.X,
ksize=[11, 11, self.CHANEELS, 96],
strides=[1, 4, 4, 1],
name='conv1')
pool1 = util.max_pool_layer(conv1,
ksize=[1, 3, 3, 1],
strides=[1, 2, 2, 1],
name='pool1')
conv2 = util.conv_layer(pool1,
ksize=[5, 5, 96, 256],
strides=[1, 1, 1, 1],
name='conv2',
padding='SAME',
group=2)
pool2 = util.max_pool_layer(conv2,
ksize=[1, 3, 3, 1],
strides=[1, 2, 2, 1],
name='pool2')
conv3 = util.conv_layer(pool2,
ksize=[3, 3, 256, 384],
strides=[1, 1, 1, 1],
name='conv3',
padding='SAME')
conv4 = util.conv_layer(conv3,
ksize=[3, 3, 384, 384],
strides=[1, 1, 1, 1],
name='conv4',
padding='SAME',
group=2)
conv5 = util.conv_layer(conv4,
ksize=[3, 3, 384, 256],
strides=[1, 1, 1, 1],
name='conv5',
padding='SAME',
group=2)
pool5 = util.max_pool_layer(conv5,
ksize=[1, 3, 3, 1],
strides=[1, 2, 2, 1],
name='pool5')
fc6 = util.full_connected_layer(pool5, 4096, name='fc6')
fc6_dropout = util.dropout(fc6, self.KEEP_PROB)
fc7 = util.full_connected_layer(fc6_dropout, 4096, name='fc7')
fc7_dropout = util.dropout(fc7, self.KEEP_PROB)
self.fc8 = util.full_connected_layer(fc7_dropout,
self.NUM_CLASSES,
name='fc8',
relu=False)
def __init__(self, optimizer, activation):
super().__init__(optimizer, activation)
############################################################################
# Define the graph #
############################################################################
# It turns out that this network from ex03 is already capable of memorizing
# the entire training or validation set, so we need to tweak generalization,
# not capacity
# In order to speed up convergence, we added batch normalization.
# Our best effort was Adam optimizer with bs=32, lr=0.001 (only possible
# because of norm)
x = tf.placeholder(tf.float32, shape=(None, 32, 32, 1), name='input')
y_ = tf.placeholder(dtype=tf.int32, shape=(None,), name='labels')
self.x = x
self.y_ = y_
kernel_shape1 = (5, 5, 1, 8)
activation1 = conv_layer(x, kernel_shape1, activation=activation)
normalize1 = batch_norm_layer(activation1)
pool1 = weighted_pool_layer(
normalize1, ksize=(1, 2, 2, 1), strides=(1, 2, 2, 1)
)
kernel_shape2 = (3, 3, 8, 10)
activation2 = conv_layer(pool1, kernel_shape2, activation=activation)
normalize2 = batch_norm_layer(activation2)
pool2 = weighted_pool_layer(
normalize2, ksize=(1, 2, 2, 1), strides=(1, 2, 2, 1)
)
pool2_reshaped = tf.reshape(pool2, (-1, 8*8*10), name='reshaped1')
fc1 = fully_connected(pool2_reshaped, 512, with_activation=True,
activation=activation)
fc2_logit = fully_connected(fc1, 10, activation=activation)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=fc2_logit,
labels=y_)
mean_cross_entropy = tf.reduce_mean(cross_entropy)
self.mean_cross_entropy = mean_cross_entropy
train_step = optimizer.minimize(mean_cross_entropy)
self.train_step = train_step
self.prediction = tf.cast(tf.argmax(fc2_logit, 1), tf.int32)
# check if neuron firing strongest coincides with max value position in real
# labels
correct_prediction = tf.equal(self.prediction, y_)
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
self.accuracy = accuracy
def create(self):
self.X = tf.reshape(self.X, shape=[-1, self.IMAGE_SIZE, self.IMAGE_SIZE, self.CHANEELS])
conv1 = util.conv_layer(self.X, ksize=[3, 3, 1, 32], strides=[1, 1, 1, 1], name='conv1')
pool1 = util.max_pool_layer(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], name='pool1')
conv2 = util.conv_layer(pool1, ksize=[3, 3, 32, 64], strides=[1, 1, 1, 1], name='conv2')
pool2 = util.max_pool_layer(conv2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], name='pool2')
conv3 = util.conv_layer(pool2, ksize=[3, 3, 64, 128], strides=[1, 1, 1, 1], name='conv3')
pool3 = util.max_pool_layer(conv3, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], name='pool3')
fc4 = util.full_connected_layer(pool3, 625, name='fc4', keep_prob=self.KEEP_PROB)
fc5 = util.full_connected_layer(fc4, 256, name='fc5', keep_prob=self.KEEP_PROB)
self.fc6 = util.full_connected_layer(fc5, 10, name='fc6')
def __init__(self, optimizer, activation):
super().__init__(optimizer, activation)
############################################################################
# Define the graph #
############################################################################
# It turns out that this network from ex03 is already capable of memorizing
# the entire training or validation set, so we need to tweak generalization,
# not capacity
x = tf.placeholder(tf.float32, shape=(None, 32, 32, 1), name='input')
y_ = tf.placeholder(dtype=tf.int32, shape=(None, ), name='labels')
self.x = x
self.y_ = y_
kernel_shape1 = (5, 5, 1, 16)
activation1 = conv_layer(x, kernel_shape1, activation=activation)
pool1 = max_pool_layer(activation1,
ksize=(1, 2, 2, 1),
strides=(1, 2, 2, 1))
kernel_shape2 = (3, 3, 16, 32)
activation2 = conv_layer(pool1, kernel_shape2, activation=activation)
pool2 = max_pool_layer(activation2,
ksize=(1, 2, 2, 1),
strides=(1, 2, 2, 1))
pool2_reshaped = tf.reshape(pool2, (-1, 2048), name='reshaped1')
fc1 = fully_connected(pool2_reshaped,
512,
with_activation=True,
activation=activation)
fc2_logit = fully_connected(fc1, 10, activation=activation)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=fc2_logit, labels=y_)
mean_cross_entropy = tf.reduce_mean(cross_entropy)
self.mean_cross_entropy = mean_cross_entropy
train_step = optimizer.minimize(mean_cross_entropy)
self.train_step = train_step
# check if neuron firing strongest coincides with max value position in real
# labels
correct_prediction = tf.equal(
tf.argmax(fc2_logit, 1, output_type=tf.int32), y_)
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
self.accuracy = accuracy
def set_model(self, figs, is_training, reuse=False):
# return only logits
h = figs
# convolution
with tf.variable_scope(self.name_scope_conv, reuse=reuse):
for i, (in_chan, out_chan) in enumerate(
zip(self.layer_chanels, self.layer_chanels[1:])):
if i == 0:
conved = conv_layer(inputs=h,
out_num=out_chan,
filter_width=5,
filter_hight=5,
stride=1,
l_id=i)
h = tf.nn.relu(conved)
#h = lrelu(conved)
else:
conved = conv_layer(inputs=h,
out_num=out_chan,
filter_width=5,
filter_hight=5,
stride=2,
l_id=i)
bn_conved = batch_norm(conved, i, is_training)
h = tf.nn.relu(bn_conved)
#h = lrelu(bn_conved)
feature_image = h
# full connect
dim = get_dim(h)
h = tf.reshape(h, [-1, dim])
with tf.variable_scope(self.name_scope_fc, reuse=reuse):
weights = get_weights('fc', [dim, self.fc_dim], 0.02)
biases = get_biases('fc', [self.fc_dim], 0.0)
h = tf.matmul(h, weights) + biases
h = batch_norm(h, 'fc', is_training)
h = tf.nn.relu(h)
weights = get_weights('fc2', [self.fc_dim, 1], 0.02)
biases = get_biases('fc2', [1], 0.0)
h = tf.matmul(h, weights) + biases
return h, feature_image
def build_model(self, image, reuse=False):
with tf.variable_scope('is_training', reuse=True):
is_training = tf.get_variable('is_training', dtype=tf.bool)
with tf.variable_scope("D_" + self.signature) as scope:
if reuse:
scope.reuse_variables()
conv_num = self.conv_infos['conv_layer_number']
conv_filter = self.conv_infos['filter']
conv_stride = self.conv_infos['stride']
prev = image
for i in range(conv_num):
if i == 0 or i == conv_num - 1:
prev = conv_layer(prev, conv_filter[i], "d_conv_{}".format(i), activation=lrelu, batch_norm=None, reuse=reuse)
else:
bn = batch_norm(name="d_bn_{}".format(i))
prev = conv_layer(prev, conv_filter[i], "d_conv_{}".format(i), activation=lrelu, batch_norm=bn, reuse=reuse)
setattr(self, "conv_{}".format(i), prev)
return tf.sigmoid(prev)
def set_model(self, figs, is_training, reuse=False):
u'''
return only logits. not sigmoid(logits).
'''
h = figs
# convolution
with tf.variable_scope(self.name_scope_conv, reuse=reuse):
for i, (in_chan, out_chan) in enumerate(
zip(self.layer_chanels, self.layer_chanels[1:])):
conved = conv_layer(inputs=h,
out_num=out_chan,
filter_width=5,
filter_hight=5,
stride=2,
l_id=i)
if i == 0:
h = tf.nn.relu(conved)
#h = lrelu(conved)
else:
bn_conved = batch_norm(conved, i, is_training)
h = tf.nn.relu(bn_conved)
#h = lrelu(bn_conved)
# full connect
dim = get_dim(h)
h = tf.reshape(h, [-1, dim])
with tf.variable_scope(self.name_scope_fc, reuse=reuse):
weights = get_weights('fc', [dim, self.fc_dim], 0.02)
biases = get_biases('fc', [self.fc_dim], 0.0)
h = tf.matmul(h, weights) + biases
h = batch_norm(h, 'en_fc_bn', is_training)
h = tf.nn.relu(h)
weights = get_weights('mu', [self.fc_dim, self.z_dim], 0.02)
biases = get_biases('mu', [self.z_dim], 0.0)
mu = tf.matmul(h, weights) + biases
weights = get_weights('sigma', [self.fc_dim, self.z_dim], 0.02)
biases = get_biases('sigma', [self.z_dim], 0.0)
log_sigma = tf.matmul(h, weights) + biases
return mu, log_sigma
def set_model(self, figs, labels, is_training, reuse=False):
fig_shape = figs.get_shape().as_list()
height, width = fig_shape[1:3]
class_num = get_dim(labels)
with tf.variable_scope(self.name_scope_label, reuse=reuse):
tmp = linear_layer(labels, class_num, height * width, 'reshape')
tmp = tf.reshape(tmp, [-1, height, width, 1])
h = tf.concat((figs, tmp), 3)
# convolution
with tf.variable_scope(self.name_scope_conv, reuse=reuse):
for i, (in_chan, out_chan) in enumerate(
zip(self.layer_chanels, self.layer_chanels[1:])):
conved = conv_layer(inputs=h,
out_num=out_chan,
filter_width=5,
filter_hight=5,
stride=2,
l_id=i)
if i == 0:
h = tf.nn.relu(conved)
#h = lrelu(conved)
else:
bn_conved = batch_norm(conved, i, is_training)
h = tf.nn.relu(bn_conved)
#h = lrelu(bn_conved)
# full connect
dim = get_dim(h)
h = tf.reshape(h, [-1, dim])
with tf.variable_scope(self.name_scope_fc, reuse=reuse):
h = linear_layer(h, dim, self.fc_dim, 'fc')
h = batch_norm(h, 'en_fc_bn', is_training)
h = tf.nn.relu(h)
mu = linear_layer(h, self.fc_dim, self.z_dim, 'mu')
log_sigma = linear_layer(h, self.fc_dim, self.z_dim, 'sigma')
return mu, log_sigma
def create(self):
self.p = []
self.X = tf.reshape(
self.X,
shape=[-1, self.IMAGE_SIZE, self.IMAGE_SIZE, self.CHANEELS])
conv1_1 = util.conv_layer(self.X,
kh=3,
kw=3,
n_out=64,
sh=1,
sw=1,
name='conv1_1',
p=self.p,
padding='SAME')
conv1_2 = util.conv_layer(conv1_1,
kh=3,
kw=3,
n_out=64,
sh=1,
sw=1,
name='conv1_2',
p=self.p,
padding='SAME')
pool1 = util.max_pool_layer(conv1_2,
kh=2,
kw=2,
sh=2,
sw=2,
name='pool1')
conv2_1 = util.conv_layer(pool1,
kh=3,
kw=3,
n_out=128,
sh=1,
sw=1,
name='conv2_1',
p=self.p,
padding='SAME')
conv2_2 = util.conv_layer(conv2_1,
kh=3,
kw=3,
n_out=128,
sh=1,
sw=1,
name='conv2_2',
p=self.p,
padding='SAME')
pool2 = util.max_pool_layer(conv2_2,
kh=2,
kw=2,
sh=2,
sw=2,
name='pool2')
conv3_1 = util.conv_layer(pool2,
kh=3,
kw=3,
n_out=256,
sh=1,
sw=1,
name='conv3_1',
p=self.p,
padding='SAME')
conv3_2 = util.conv_layer(conv3_1,
kh=3,
kw=3,
n_out=256,
sh=1,
sw=1,
name='conv3_2',
p=self.p,
padding='SAME')
conv3_3 = util.conv_layer(conv3_2,
kh=3,
kw=3,
n_out=256,
sh=1,
sw=1,
name='conv3_3',
p=self.p,
padding='SAME')
pool3 = util.max_pool_layer(conv3_3,
kh=2,
kw=2,
sh=2,
sw=2,
name='pool3')
conv4_1 = util.conv_layer(pool3,
kh=3,
kw=3,
n_out=512,
sh=1,
sw=1,
name='conv4_1',
p=self.p,
padding='SAME')
conv4_2 = util.conv_layer(conv4_1,
kh=3,
kw=3,
n_out=512,
sh=1,
sw=1,
name='conv4_2',
p=self.p,
padding='SAME')
conv4_3 = util.conv_layer(conv4_2,
kh=3,
kw=3,
n_out=512,
sh=1,
sw=1,
name='conv4_3',
p=self.p,
padding='SAME')
pool4 = util.max_pool_layer(conv4_3,
kh=2,
kw=2,
sh=2,
sw=2,
name='pool4')
conv5_1 = util.conv_layer(pool4,
kh=3,
kw=3,
n_out=512,
sh=1,
sw=1,
name='conv5_1',
p=self.p,
padding='SAME')
conv5_2 = util.conv_layer(conv5_1,
kh=3,
kw=3,
n_out=512,
sh=1,
sw=1,
name='conv5_2',
p=self.p,
padding='SAME')
conv5_3 = util.conv_layer(conv5_2,
kh=3,
kw=3,
n_out=512,
sh=1,
sw=1,
name='conv5_3',
p=self.p,
padding='SAME')
pool5 = util.max_pool_layer(conv5_3,
kh=2,
kw=2,
sh=2,
sw=2,
name='pool5')
fc6 = util.full_connected_layer(pool5,
n_out=4096,
name='fc6',
p=self.p)
fc6_dropout = util.dropout(fc6, self.KEEP_PROB, name='fc6_drop')
fc7 = util.full_connected_layer(fc6_dropout,
n_out=4096,
name='fc7',
p=self.p)
fc7_dropout = util.dropout(fc7, self.KEEP_PROB, name='fc7_drop')
self.fc8 = util.full_connected_layer(fc7_dropout,
self.NUM_CLASSES,
name='fc8',
p=self.p)
self.softmax = tf.nn.softmax(self.fc8)
self.predictions = tf.argmax(self.softmax, 1)
def __init__(self, optimizer, activation):
super().__init__(optimizer, activation)
x = tf.placeholder(tf.float32, shape=(None, 32, 32, 1), name='input')
y_ = tf.placeholder(dtype=tf.int32, shape=(None, ), name='labels')
self.x = x
self.y_ = y_
# logits = (LinearWrap(image)
# .Conv2D('conv1', 24, 5, padding='VALID')
# .MaxPooling('pool1', 2, padding='SAME')
# .Conv2D('conv2', 32, 3, padding='VALID')
# .Conv2D('conv3', 32, 3, padding='VALID')
# .MaxPooling('pool2', 2, padding='SAME')
# .Conv2D('conv4', 64, 3, padding='VALID')
# .Dropout('drop', 0.5)
# .FullyConnected('fc0', 512,
# b_init=tf.constant_initializer(0.1), nl=tf.nn.relu)
# .FullyConnected('linear', out_dim=10, nl=tf.identity)())
# tf.nn.softmax(logits, name='output')
l1 = conv_layer(x, (5, 5, 1, 24),
activation=None,
padding='VALID',
use_bias=False)
l2 = tf.nn.relu(batch_norm_layer(l1))
l3 = tf.nn.max_pool(l2, (1, 2, 2, 1), (1, 2, 2, 1), padding='SAME')
l4 = conv_layer(l3, (3, 3, 24, 32),
padding='VALID',
activation=None,
use_bias=False)
l5 = tf.nn.relu(batch_norm_layer(l4))
l6 = conv_layer(l5, (3, 3, 32, 32),
padding='VALID',
activation=None,
use_bias=False)
l7 = tf.nn.relu(batch_norm_layer(l6))
l8 = tf.nn.max_pool(l7, (1, 2, 2, 1), (1, 2, 2, 1), padding='SAME')
l8_ = conv_layer(l8, (3, 3, 32, 64),
padding='VALID',
activation=None,
use_bias=False)
l9 = tf.nn.relu(batch_norm_layer(l8_))
l10 = tf.nn.dropout(l9, 0.5)
l11 = tf.reshape(l10, (-1, 3 * 3 * 64), name='reshaped1')
l12 = fully_connected(l11,
512,
with_activation=True,
activation=tf.nn.relu)
l13 = fully_connected(l12, 10, with_activation=False, use_bias=False)
cross_entropy = tf.nn.sparse_softmax_cross_entropy_with_logits(
logits=l13, labels=y_)
mean_cross_entropy = tf.reduce_mean(cross_entropy)
self.mean_cross_entropy = mean_cross_entropy
train_step = optimizer.minimize(mean_cross_entropy)
self.train_step = train_step
self.prediction = tf.cast(tf.argmax(l13, 1), tf.int32)
# check if neuron firing strongest coincides with max value position in real
# labels
correct_prediction = tf.equal(self.prediction, y_)
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
self.accuracy = accuracy
