def build_network():
network = resnet_model.resnet_v1(
resnet_depth=params['resnet_depth'],
num_classes=params['num_label_classes'],
dropblock_size=params['dropblock_size'],
dropblock_keep_probs=dropblock_keep_probs,
data_format=params['data_format'])
return network(inputs=features,
is_training=(mode == tf.estimator.ModeKeys.TRAIN))
def build_network():
network = resnet_model.resnet_v1(
resnet_depth=FLAGS.resnet_depth,
num_classes=FLAGS.num_label_classes,
dropblock_size=FLAGS.dropblock_size,
dropblock_keep_probs=dropblock_keep_probs,
data_format=FLAGS.data_format)
return network(inputs=features,
is_training=(mode == tf.estimator.ModeKeys.TRAIN))
def test_load_resnet18_v1(self):
network = resnet_model.resnet_v1(resnet_depth=18,
num_classes=10,
data_format='channels_last')
input_bhw3 = tf.placeholder(tf.float32, [1, 28, 28, 3])
resnet_output = network(inputs=input_bhw3, is_training=True)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
_ = sess.run(resnet_output,
feed_dict={input_bhw3: np.random.randn(1, 28, 28, 3)})
def test_load_resnet_rs(self):
network = resnet_model.resnet_v1(resnet_depth=50,
num_classes=10,
data_format='channels_last',
se_ratio=0.25,
drop_connect_rate=0.2,
use_resnetd_stem=True,
resnetd_shortcut=True,
dropout_rate=0.2,
replace_stem_max_pool=True)
input_bhw3 = tf.placeholder(tf.float32, [1, 28, 28, 3])
resnet_output = network(inputs=input_bhw3, is_training=True)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
_ = sess.run(resnet_output,
feed_dict={input_bhw3: np.random.randn(1, 28, 28, 3)})
