diff --git a/resnet.py b/resnet.py
index a056c41..7270abe 100644
--- a/resnet.py
+++ b/resnet.py
@@ -101,7 +101,7 @@ def inference_small(x,
     c['fc_units_out'] = num_classes
     c['num_blocks'] = num_blocks
     c['num_classes'] = num_classes
-    inference_small_config(x, c)
+    return inference_small_config(x, c)
 
 def inference_small_config(x, c):
     c['bottleneck'] = False
@@ -151,7 +151,7 @@ def loss(logits, labels):
     regularization_losses = tf.get_collection(tf.GraphKeys.REGULARIZATION_LOSSES)
 
     loss_ = tf.add_n([cross_entropy_mean] + regularization_losses)
-    tf.scalar_summary('loss', loss_)
+    tf.summary.scalar('loss', loss_)
 
     return loss_
 
@@ -241,7 +241,7 @@ def bn(x, c):
                          initializer=tf.zeros_initializer)
     gamma = _get_variable('gamma',
                           params_shape,
-                          initializer=tf.ones_initializer)
+                          initializer=tf.ones_initializer())
 
     moving_mean = _get_variable('moving_mean',
                                 params_shape,
@@ -249,7 +249,7 @@ def bn(x, c):
                                 trainable=False)
     moving_variance = _get_variable('moving_variance',
                                     params_shape,
-                                    initializer=tf.ones_initializer,
+                                    initializer=tf.ones_initializer(),
                                     trainable=False)
 
     # These ops will only be preformed when training.
@@ -300,7 +300,7 @@ def _get_variable(name,
         regularizer = tf.contrib.layers.l2_regularizer(weight_decay)
     else:
         regularizer = None
-    collections = [tf.GraphKeys.VARIABLES, RESNET_VARIABLES]
+    collections = [tf.GraphKeys.GLOBAL_VARIABLES, RESNET_VARIABLES]
     return tf.get_variable(name,
                            shape=shape,
                            initializer=initializer,
diff --git a/resnet_train.py b/resnet_train.py
index ce6a0bb..8111a7b 100644
--- a/resnet_train.py
+++ b/resnet_train.py
@@ -39,15 +39,15 @@ def train(is_training, logits, images, labels):
     # loss_avg
     ema = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)
     tf.add_to_collection(UPDATE_OPS_COLLECTION, ema.apply([loss_]))
-    tf.scalar_summary('loss_avg', ema.average(loss_))
+    tf.summary.scalar('loss_avg', ema.average(loss_))
 
     # validation stats
     ema = tf.train.ExponentialMovingAverage(0.9, val_step)
     val_op = tf.group(val_step.assign_add(1), ema.apply([top1_error]))
     top1_error_avg = ema.average(top1_error)
-    tf.scalar_summary('val_top1_error_avg', top1_error_avg)
+    tf.summary.scalar('val_top1_error_avg', top1_error_avg)
 
-    tf.scalar_summary('learning_rate', FLAGS.learning_rate)
+    tf.summary.scalar('learning_rate', FLAGS.learning_rate)
 
     opt = tf.train.MomentumOptimizer(FLAGS.learning_rate, MOMENTUM)
     grads = opt.compute_gradients(loss_)
@@ -67,17 +67,17 @@ def train(is_training, logits, images, labels):
     batchnorm_updates_op = tf.group(*batchnorm_updates)
     train_op = tf.group(apply_gradient_op, batchnorm_updates_op)
 
-    saver = tf.train.Saver(tf.all_variables())
+    saver = tf.train.Saver(tf.global_variables())
 
-    summary_op = tf.merge_all_summaries()
+    summary_op = tf.summary.merge_all()
 
-    init = tf.initialize_all_variables()
+    init = tf.global_variables_initializer()
 
     sess = tf.Session(config=tf.ConfigProto(log_device_placement=False))
     sess.run(init)
     tf.train.start_queue_runners(sess=sess)
 
-    summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph)
+    summary_writer = tf.summary.FileWriter(FLAGS.train_dir, sess.graph)
 
     if FLAGS.resume:
         latest = tf.train.latest_checkpoint(FLAGS.train_dir)
diff --git a/train_cifar.py b/train_cifar.py
index cccf590..80391de 100644
--- a/train_cifar.py
+++ b/train_cifar.py
@@ -192,7 +192,7 @@ def distorted_inputs(data_dir, batch_size):
         distorted_image, lower=0.2, upper=1.8)
 
     # Subtract off the mean and divide by the variance of the pixels.
-    float_image = tf.image.per_image_whitening(distorted_image)
+    float_image = tf.image.per_image_standardization(distorted_image)
 
     # Ensure that the random shuffling has good mixing properties.
     min_fraction_of_examples_in_queue = 0.4
@@ -250,7 +250,7 @@ def inputs(eval_data, data_dir, batch_size):
                                                            width, height)
 
     # Subtract off the mean and divide by the variance of the pixels.
-    float_image = tf.image.per_image_whitening(resized_image)
+    float_image = tf.image.per_image_standardization(resized_image)
 
     # Ensure that the random shuffling has good mixing properties.
     min_fraction_of_examples_in_queue = 0.4