def mt_loop(self, net, reuse=False):
bn_params = {
'decay': 0.999,
'center': True,
'scale': True,
'epsilon': 0.001,
'updates_collections': None,
'is_training': self.is_training
}
with tf.variable_scope('mt_loop', reuse=reuse):
net = layers.conv2d(
net,
48,
kernel_size=3,
name='conv1',
reuse=reuse,
stride=2,
padding='valid',
normalizer_fn=layers.batchNormalization,
normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool2d(net, 3, 2, name='pool1')
net = layers.squeeze_and_excite2d(net,
indexHeight=1,
indexWidth=2,
name='se1',
filters=48,
reuse=reuse)
return net
def build_resnet_18(self, net, reuse=False):
bn_params = {
'decay': 0.999,
'center': True,
'scale': True,
'epsilon': 0.001,
'updates_collections': None,
'is_training': self.is_training
}
net = layers.conv2d(
net,
filters=64,
kernel_size=7,
stride=2,
padding='SAME',
name='resnet_conv1',
reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization,
normalizer_params=bn_params)
net = layers.max_pool2d(net, kernel_size=3, stride=2, padding='same')
net = layers.squeeze_and_excite2d(net,
indexHeight=1,
indexWidth=2,
name='se1',
filters=64)
net = self.res_block(net, 64, 3, reuse=reuse, name='resnet_conv2a')
net = self.res_block(net, 64, 3, reuse=reuse, name='resnet_conv2b')
net = self.res_block_first(net,
128,
3,
reuse=reuse,
name='resnet_conv3a')
net = self.res_block(net, 128, 3, reuse=reuse, name='resnet_conv3b')
net = self.res_block_first(net,
256,
3,
reuse=reuse,
name='resnet_conv4a')
net = self.res_block(net, 256, 3, reuse=reuse, name='resnet_conv4b')
net = self.res_block_first(net,
512,
3,
reuse=reuse,
name='resnet_conv5a')
net = self.res_block(net, 512, 3, reuse=reuse, name='resnet_conv5b')
net = layers.avg_pool2d(net, kernel_size=7, stride=1)
return net
def mt_loop(self, net, reuse=False):
bn_params = {'decay': 0.999, 'center': True, 'scale': True, 'epsilon': 0.001, 'updates_collections': None,
'is_training': self.is_training}
with tf.variable_scope('mt_loop', reuse=reuse):
net = layers.conv2d(net, 48, kernel_size=3, name='conv1', reuse=reuse, stride=2, padding='valid',
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool2d(net, 3, 2, name='pool1')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se1', filters=48, reuse=reuse)
return net
def res_block(self,
net,
filters=64,
kernel=3,
name='resnet_convX',
reuse=False):
bn_params = {
'decay': 0.999,
'center': True,
'scale': True,
'epsilon': 0.001,
'updates_collections': None,
'is_training': self.is_training
}
tmp_logits = net
net = layers.conv2d(
net,
filters=filters,
kernel_size=kernel,
stride=1,
padding='same',
name=name + '_1',
reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization,
normalizer_params=bn_params)
net = layers.conv2d(
net,
filters=filters,
kernel_size=kernel,
stride=1,
padding='same',
name=name + '_2',
reuse=reuse,
activation_fn=None,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization,
normalizer_params=bn_params)
net = layers.squeeze_and_excite2d(net,
indexHeight=1,
indexWidth=2,
name='se' + name,
filters=filters)
net = layers.add(net, tmp_logits)
net = layers.relu(net)
return net
def createModel(self):
print("CREATING MODEL")
self.global_step = tf.Variable(0, trainable=False)
self.is_training = tf.placeholder_with_default(True, [], name='is_training')
self.learning_rate = layers.decayLearningRate(LEARNING_RATE, self.global_step, DECAY_STEPS, DECAY_RATE)
self.X = tf.placeholder(dtype=tf.float32, shape=[None, self.dataset.frames, self.dataset.h, self.dataset.w, self.dataset.c])
self.Y = tf.placeholder(dtype=tf.int32, shape=[None])
self.Yoh = layers.toOneHot(self.Y, self.dataset.num_classes)
reuse = None
towersLogits = []
for sequence_image in range(self.dataset.frames):
net = self.vgg16(self.X[:, sequence_image], reuse)
towersLogits.append(net)
reuse = True
net = layers.stack(towersLogits)
del towersLogits[:]
net = layers.transpose(net, [1, 2, 3, 0, 4])
net = layers.reshape(net, [-1, net.shape[1], net.shape[2], net.shape[3] * net.shape[4]])
net = layers.fc(net, 512, name='fc5', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se5', filters=512)
net = layers.flatten(net, name='flatten')
net = layers.fc(net, 4096, name='fc6', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.fc(net, 4096, name='fc7', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
self.logits = layers.fc(net, self.dataset.num_classes, activation_fn=None, name='fc8',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
self.preds = layers.softmax(self.logits)
cross_entropy_loss = layers.reduce_mean(layers.softmax_cross_entropy(logits=self.logits, labels=self.Yoh))
regularization_loss = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)
self.loss = cross_entropy_loss + REGULARIZER_SCALE * tf.reduce_sum(regularization_loss)
self.opt = layers.adam(self.learning_rate)
self.train_op = self.opt.minimize(self.loss, global_step=self.global_step)
self.accuracy, self.precision, self.recall = self.createSummaries(self.Yoh, self.preds, self.loss,
self.learning_rate)
def modelLogits(self):
self.towerLogits = tf.placeholder(dtype=tf.float32, shape=[None, self.dataset.frames, 3, 3, 512])
net = self.towerLogits
net = layers.transpose(net, [0, 2, 3, 1, 4])
net = layers.reshape(net, [-1, net.shape[1], net.shape[2], net.shape[3] * net.shape[4]])
net = layers.fc(net, 512, name='fc5', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se5', filters=512)
net = layers.flatten(net, name='flatten')
net = layers.fc(net, 4096, name='fc6', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.fc(net, 4096, name='fc7', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
self.logits = layers.fc(net, self.dataset.num_classes, activation_fn=None, name='fc8',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
def createModel(self):
print("CREATING MODEL")
self.global_step = tf.Variable(0, trainable=False)
self.is_training = tf.placeholder_with_default(True, [], name='is_training')
self.learning_rate = layers.decayLearningRate(LEARNING_RATE, self.global_step, DECAY_STEPS, DECAY_RATE)
self.X = tf.placeholder(dtype=tf.float32, shape=[None, self.dataset.frames, self.dataset.h, self.dataset.w, self.dataset.c])
self.Y = tf.placeholder(dtype=tf.int32, shape=[None])
self.Yoh = layers.toOneHot(self.Y, self.dataset.num_classes)
net = self.X
if net.shape[-1] > 1:
net = layers.rgb_to_grayscale(net)
if len(net.shape) > 4:
net = tf.transpose(net, [0, 2, 3, 1, 4])
net = tf.reshape(net, [-1, net.shape[1], net.shape[2], net.shape[3] * net.shape[4]])
bn_params = {'decay': 0.999, 'center': True, 'scale': True, 'epsilon': 0.001,
'updates_collections': None, 'is_training': self.is_training}
net = layers.conv2d(net, filters=96, kernel_size=3, padding='VALID', stride=2, name='conv1',
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool2d(net, 3, 2, name='max_pool1')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se1', filters=96)
net = layers.conv2d(net, filters=256, kernel_size=3, padding='VALID', stride=2, name='conv2',
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool2d(net, 3, 2, name='max_pool2')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se2', filters=256)
net = layers.conv2d(net, filters=512, kernel_size=3, padding='SAME', stride=1, name='conv3',
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se3', filters=512)
net = layers.conv2d(net, filters=512, kernel_size=3, padding='SAME', stride=1, name='conv4',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se4', filters=512)
net = layers.conv2d(net, filters=512, kernel_size=3, padding='SAME', stride=1, name='conv5',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool2d(net, 3, 2, name='max_pool2')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='se5', filters=512)
net = layers.flatten(net, name='flatten')
net = layers.fc(net, 4096, name='fc6', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.fc(net, 4096, name='fc7', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
self.logits = layers.fc(net, self.dataset.num_classes, activation_fn=None, name='fc8', weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
self.preds = layers.softmax(self.logits)
cross_entropy_loss = layers.reduce_mean(layers.softmax_cross_entropy(logits=self.logits, labels=self.Yoh))
regularization_loss = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)
self.loss = cross_entropy_loss + REGULARIZER_SCALE * tf.reduce_sum(regularization_loss)
self.opt = layers.adam(self.learning_rate)
self.train_op = self.opt.minimize(self.loss, global_step=self.global_step)
self.accuracy, self.precision, self.recall = self.createSummaries(self.Yoh, self.preds, self.loss,
self.learning_rate)
def vgg(self, net, reuse=False):
with tf.variable_scope('vgg_16', reuse=reuse):
bn_params = {'decay': 0.999, 'center': True, 'scale': True, 'epsilon': 0.001, 'updates_collections': None,
'is_training': self.is_training}
with tf.variable_scope('conv1', reuse=reuse):
net = layers.conv2d(net, 64, name='conv1_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 64, name='conv1_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool1')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se1', filters=64)
with tf.variable_scope('conv2', reuse=reuse):
net = layers.conv2d(net, 128, name='conv2_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 128, name='conv2_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool2')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se2', filters=128)
with tf.variable_scope('conv3', reuse=reuse):
net = layers.conv2d(net, 256, name='conv3_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 256, name='conv3_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 256, name='conv3_3', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool3')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se3', filters=256)
with tf.variable_scope('conv4', reuse=reuse):
net = layers.conv2d(net, 512, name='conv4_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv4_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv4_3', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool4')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se4', filters=512)
with tf.variable_scope('conv5', reuse=reuse):
net = layers.conv2d(net, 512, name='conv5_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv5_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv5_3', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool5')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se5', filters=512)
return net
def createModel(self):
print("CREATING MODEL")
self.global_step = tf.Variable(0, trainable=False)
self.is_training = tf.placeholder_with_default(True, [],
name='is_training')
self.learning_rate = layers.decayLearningRate(LEARNING_RATE,
self.global_step,
DECAY_STEPS, DECAY_RATE)
self.X = tf.placeholder(dtype=tf.float32,
shape=[
None, self.dataset.frames, self.dataset.h,
self.dataset.w, self.dataset.c
])
self.Y = tf.placeholder(dtype=tf.int32, shape=[None])
self.Yoh = layers.toOneHot(self.Y, self.dataset.num_classes)
net = self.X
conv3d_kernel = [3, 3, 3]
max3d_pool_kernel = [3, 3, 2]
bn_params = {
'decay': 0.999,
'center': True,
'scale': True,
'epsilon': 0.001,
'updates_collections': None,
'is_training': self.is_training
}
net = layers.transpose(net, [0, 2, 3, 1, 4])
net = layers.conv3d(
net,
filters=48,
kernel_size=conv3d_kernel,
padding='VALID',
stride=2,
name='conv1',
normalizer_fn=layers.batchNormalization,
normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool3d(net,
max3d_pool_kernel,
2,
padding='VALID',
name='max_pool1')
net = layers.squeeze_and_excite3d(net,
indexHeight=1,
indexWidth=2,
indexSeq=3,
name='se1',
filters=48)
net = layers.conv3d(
net,
filters=256,
kernel_size=conv3d_kernel,
padding='VALID',
stride=2,
name='conv2',
normalizer_fn=layers.batchNormalization,
normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool3d(net,
max3d_pool_kernel,
2,
padding='VALID',
name='max_pool2')
net = layers.squeeze_and_excite3d(net,
indexHeight=1,
indexWidth=2,
indexSeq=3,
name='se2',
filters=256)
net = layers.reshape(
net, [-1, net.shape[1], net.shape[2], net.shape[3] * net.shape[4]])
net = layers.conv2d(
net,
filters=512,
kernel_size=3,
padding='SAME',
stride=1,
name='conv3',
normalizer_fn=layers.batchNormalization,
normalizer_params=bn_params,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.squeeze_and_excite2d(net,
indexHeight=1,
indexWidth=2,
name='se3',
filters=512)
net = layers.conv2d(
net,
filters=512,
kernel_size=3,
padding='SAME',
stride=1,
name='conv4',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.squeeze_and_excite2d(net,
indexHeight=1,
indexWidth=2,
name='se4',
filters=512)
net = layers.conv2d(
net,
filters=512,
kernel_size=3,
padding='SAME',
stride=1,
name='conv5',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.max_pool2d(net, 3, 1, padding='VALID', name='max_pool5')
net = layers.squeeze_and_excite2d(net,
indexHeight=1,
indexWidth=2,
name='se5',
filters=512)
net = layers.flatten(net, name='flatten')
net = layers.fc(
net,
4096,
name='fc6',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
net = layers.fc(
net,
4096,
name='fc7',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
self.logits = layers.fc(
net,
self.dataset.num_classes,
activation_fn=None,
name='fc8',
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE))
self.preds = layers.softmax(self.logits)
cross_entropy_loss = layers.reduce_mean(
layers.softmax_cross_entropy(logits=self.logits, labels=self.Yoh))
regularization_loss = tf.get_collection(
tf.GraphKeys.REGULARIZATION_LOSSES)
self.loss = cross_entropy_loss + REGULARIZER_SCALE * tf.reduce_sum(
regularization_loss)
self.opt = layers.adam(self.learning_rate)
self.train_op = self.opt.minimize(self.loss,
global_step=self.global_step)
self.accuracy, self.precision, self.recall = self.createSummaries(
self.Yoh, self.preds, self.loss, self.learning_rate)
def vgg16(self, net, reuse=False):
with tf.variable_scope('vgg_16', reuse=reuse):
bn_params = {'decay': 0.999, 'center': True, 'scale': True, 'epsilon': 0.001, 'updates_collections': None,
'is_training': self.is_training}
with tf.variable_scope('conv1', reuse=reuse):
net = layers.conv2d(net, 64, name='conv1_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 64, name='conv1_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool1')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se1', filters=64)
with tf.variable_scope('conv2', reuse=reuse):
net = layers.conv2d(net, 128, name='conv2_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 128, name='conv2_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool2')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se2', filters=128)
with tf.variable_scope('conv3', reuse=reuse):
net = layers.conv2d(net, 256, name='conv3_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 256, name='conv3_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 256, name='conv3_3', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool3')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se3', filters=256)
with tf.variable_scope('conv4', reuse=reuse):
net = layers.conv2d(net, 512, name='conv4_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv4_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv4_3', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool4')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se4', filters=512)
with tf.variable_scope('conv5', reuse=reuse):
net = layers.conv2d(net, 512, name='conv5_1', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv5_2', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.conv2d(net, 512, name='conv5_3', reuse=reuse,
weights_regularizer=layers.l2_regularizer(REGULARIZER_SCALE),
normalizer_fn=layers.batchNormalization, normalizer_params=bn_params)
net = layers.max_pool2d(net, [2, 2], 2, name='pool5')
net = layers.squeeze_and_excite2d(net, indexHeight=1, indexWidth=2, name='vgg_se5', filters=512)
return net