def testRaiseValueErrorWithInvalidDepthMultiplier(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
with self.assertRaises(ValueError):
_ = inception_v2.inception_v2(inputs, num_classes, depth_multiplier=-0.1)
with self.assertRaises(ValueError):
_ = inception_v2.inception_v2(inputs, num_classes, depth_multiplier=0.0)
def testRaiseValueErrorWithInvalidDepthMultiplier(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
with self.assertRaises(ValueError):
_ = inception_v2.inception_v2(inputs,
num_classes,
depth_multiplier=-0.1)
with self.assertRaises(ValueError):
_ = inception_v2.inception_v2(inputs,
num_classes,
depth_multiplier=0.0)
def network_inception_v2():
input_shape = [1, 224, 224, 3]
input_ = tf.placeholder(dtype=tf.float32, name='input', shape=input_shape)
net, _end_points = inception_v2(input_,
num_classes=1000,
is_training=False)
return net
def testTrainEvalWithReuse(self):
train_batch_size = 5
eval_batch_size = 2
height, width = 150, 150
num_classes = 1000
train_inputs = random_ops.random_uniform(
(train_batch_size, height, width, 3))
inception_v2.inception_v2(train_inputs, num_classes)
eval_inputs = random_ops.random_uniform((eval_batch_size, height, width, 3))
logits, _ = inception_v2.inception_v2(eval_inputs, num_classes, reuse=True)
predictions = math_ops.argmax(logits, 1)
with self.test_session() as sess:
sess.run(variables.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (eval_batch_size,))
def inception_v2_layer(self):
out_dimension = self.w_len
with slim.arg_scope(inception_v2_arg_scope()):
out, end_points = inception_v2(inputs=self.img_input_expand,
num_classes=out_dimension,
dropout_keep_prob=self.keep_prob,
is_training=self.is_training)
return out
def testLogitsNotSqueezed(self):
num_classes = 25
images = random_ops.random_uniform([1, 224, 224, 3])
logits, _ = inception_v2.inception_v2(
images, num_classes=num_classes, spatial_squeeze=False)
with self.test_session() as sess:
variables.global_variables_initializer().run()
logits_out = sess.run(logits)
self.assertListEqual(list(logits_out.shape), [1, 1, 1, num_classes])
def _build(self, inputs, gt_embedding):
scope = tf.get_variable_scope()
scope.set_initializer(
tf.random_uniform_initializer(-self.params["init_scale"],
self.params["init_scale"]))
max_input_seq_len = self.params["max_input_seq_length"]
input_embeding_dim = self.params["input_embeding_dim"]
batch_size = tf.shape(inputs)[0]
# Patch or trim dynamic input to fixed size.
# inputs data is RNN result in shape of [batch, seq_len, embedding_dim]
# Patch with 0 or trim extra length and make the tensor shape to be
# [batch, max_input_seq_len, embedding_dim]
target_size = ops.convert_to_tensor(max_input_seq_len,
dtype=dtypes.int32,
name="max_input_seq_len")
def _pad_fn(inputs, max_input_seq_len):
target_pad = max_input_seq_len - tf.shape(inputs)[1]
# inputs = tf.Print(inputs, [tf.shape(inputs), tf.transpose(inputs, [0, 2, 1])], message="input before pad: ", summarize = 1000)
padded_inputs = tf.pad(inputs, [[0, 0], [0, target_pad], [0, 0]],
"CONSTANT")
# padded_inputs = tf.Print(padded_inputs, [tf.shape(padded_inputs), tf.transpose(padded_inputs, [0, 2, 1])], message="input after pad: ", summarize = 1000)
return padded_inputs
pad_inputs = tf.cond(tf.shape(inputs)[1] < max_input_seq_len,
lambda: _pad_fn(inputs, max_input_seq_len),
lambda: tf.identity(inputs),
name="maybe_padding")
def _trim_fn(inputs, max_input_seq_len):
return inputs
final_inputs = tf.cond(tf.shape(pad_inputs)[1] > max_input_seq_len,
lambda: _trim_fn(pad_inputs, max_input_seq_len),
lambda: tf.identity(pad_inputs),
name="maybe_trimming")
# Inception input must be rank 4: [batch, row, col, channel], we have to
# manually expand our rank3 input.
inception_input = tf.expand_dims(final_inputs, 3)
# Input size for inception must be 224*224
inception_input = tf.image.resize_images(
images=inception_input,
size=[224, 224],
method=tf.image.ResizeMethod.BILINEAR)
logits, _ = inception_v2(inputs=inception_input,
num_classes=self.vocab_size,
is_training=True)
predictions = math_ops.cast(math_ops.argmax(logits, axis=-1),
dtypes.int32)
return DecoderOutput(logits=logits, predicted_ids=predictions), None
def testTrainEvalWithReuse(self):
train_batch_size = 5
eval_batch_size = 2
height, width = 150, 150
num_classes = 1000
train_inputs = random_ops.random_uniform(
(train_batch_size, height, width, 3))
inception_v2.inception_v2(train_inputs, num_classes)
eval_inputs = random_ops.random_uniform(
(eval_batch_size, height, width, 3))
logits, _ = inception_v2.inception_v2(eval_inputs,
num_classes,
reuse=True)
predictions = math_ops.argmax(logits, 1)
with self.test_session() as sess:
sess.run(variables.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (eval_batch_size, ))
def testBuildEndPointsWithDepthMultiplierGreaterThanOne(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
_, end_points = inception_v2.inception_v2(inputs, num_classes)
endpoint_keys = [
key for key in end_points.keys()
if key.startswith('Mixed') or key.startswith('Conv')
]
_, end_points_with_multiplier = inception_v2.inception_v2(
inputs, num_classes, scope='depth_multiplied_net', depth_multiplier=2.0)
for key in endpoint_keys:
original_depth = end_points[key].get_shape().as_list()[3]
new_depth = end_points_with_multiplier[key].get_shape().as_list()[3]
self.assertEqual(2.0 * original_depth, new_depth)
def testLogitsNotSqueezed(self):
num_classes = 25
images = random_ops.random_uniform([1, 224, 224, 3])
logits, _ = inception_v2.inception_v2(images,
num_classes=num_classes,
spatial_squeeze=False)
with self.test_session() as sess:
variables.global_variables_initializer().run()
logits_out = sess.run(logits)
self.assertListEqual(list(logits_out.shape),
[1, 1, 1, num_classes])
def testHalfSizeImages(self):
batch_size = 5
height, width = 112, 112
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
logits, end_points = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
pre_pool = end_points['Mixed_5c']
self.assertListEqual(pre_pool.get_shape().as_list(),
[batch_size, 4, 4, 1024])
def testBuildClassificationNetwork(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
logits, end_points = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
self.assertTrue('Predictions' in end_points)
self.assertListEqual(end_points['Predictions'].get_shape().as_list(),
[batch_size, num_classes])
def testHalfSizeImages(self):
batch_size = 5
height, width = 112, 112
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
logits, end_points = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
pre_pool = end_points['Mixed_5c']
self.assertListEqual(pre_pool.get_shape().as_list(),
[batch_size, 4, 4, 1024])
def testBuildClassificationNetwork(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
logits, end_points = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
self.assertTrue('Predictions' in end_points)
self.assertListEqual(end_points['Predictions'].get_shape().as_list(),
[batch_size, num_classes])
def inference(self):
x = tf.reshape(self.x,
shape=[
-1, self.input_shape[0], self.input_shape[1],
self.input_shape[2]
])
with slim.arg_scope(inception_v2.inception_v2_arg_scope()):
logits, end_points = inception_v2.inception_v2(
x,
num_classes=self.nclasses,
is_training=self.is_training,
spatial_squeeze=True)
return logits
def testEvaluation(self):
batch_size = 2
height, width = 224, 224
num_classes = 1000
eval_inputs = random_ops.random_uniform((batch_size, height, width, 3))
logits, _ = inception_v2.inception_v2(
eval_inputs, num_classes, is_training=False)
predictions = math_ops.argmax(logits, 1)
with self.test_session() as sess:
sess.run(variables.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size,))
def testBuildEndPointsWithDepthMultiplierGreaterThanOne(self):
batch_size = 5
height, width = 224, 224
num_classes = 1000
inputs = random_ops.random_uniform((batch_size, height, width, 3))
_, end_points = inception_v2.inception_v2(inputs, num_classes)
endpoint_keys = [
key for key in end_points.keys()
if key.startswith('Mixed') or key.startswith('Conv')
]
_, end_points_with_multiplier = inception_v2.inception_v2(
inputs,
num_classes,
scope='depth_multiplied_net',
depth_multiplier=2.0)
for key in endpoint_keys:
original_depth = end_points[key].get_shape().as_list()[3]
new_depth = end_points_with_multiplier[key].get_shape().as_list(
)[3]
self.assertEqual(2.0 * original_depth, new_depth)
def testUnknownBatchSize(self):
batch_size = 1
height, width = 224, 224
num_classes = 1000
inputs = array_ops.placeholder(dtypes.float32, (None, height, width, 3))
logits, _ = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(), [None, num_classes])
images = random_ops.random_uniform((batch_size, height, width, 3))
with self.test_session() as sess:
sess.run(variables.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()})
self.assertEquals(output.shape, (batch_size, num_classes))
def testEvaluation(self):
batch_size = 2
height, width = 224, 224
num_classes = 1000
eval_inputs = random_ops.random_uniform((batch_size, height, width, 3))
logits, _ = inception_v2.inception_v2(eval_inputs,
num_classes,
is_training=False)
predictions = math_ops.argmax(logits, 1)
with self.test_session() as sess:
sess.run(variables.global_variables_initializer())
output = sess.run(predictions)
self.assertEquals(output.shape, (batch_size, ))
def test_program():
with tf.Graph().as_default():
slim = tf.contrib.slim
checkpoint_dir = '/home/taivu/workspace/NudityDetection/Trained_weight'
input_tensor = tf.placeholder(tf.float32,
shape=(None, 224, 224, 3),
name='input_image')
scaled_input_tensor = tf.scalar_mul((1.0 / 255), input_tensor)
scaled_input_tensor = tf.sub(scaled_input_tensor, 0.5)
scaled_input_tensor = tf.mul(scaled_input_tensor, 2.0)
arg_scope = inception_v2_arg_scope()
with slim.arg_scope(arg_scope):
logits, end_points = inception_v2(scaled_input_tensor,
is_training=False)
init = tf.global_variables_initializer()
saver = tf.train.Saver(tf.global_variables())
coord = tf.train.Coordinator()
with tf.Session() as sess:
sess.run(init)
threads = tf.train.start_queue_runners(sess, coord)
ckpt = tf.train.get_checkpoint_state(checkpoint_dir)
if ckpt and ckpt.model_checkpoint_path:
saver.restore(sess, ckpt.model_checkpoint_path)
im = Image.open(
'/home/taivu/workspace/NudityDetection/dog.jpeg').resize(
(224, 224))
im = np.array(im)
im = im.reshape(1, 224, 224, 3)
predict_values, logit_values = sess.run(
[end_points['Predictions'], logits],
feed_dict={input_tensor: im})
print(np.max(predict_values), np.max(logit_values))
coord.request_stop()
def testUnknownBatchSize(self):
batch_size = 1
height, width = 224, 224
num_classes = 1000
inputs = array_ops.placeholder(dtypes.float32,
(None, height, width, 3))
logits, _ = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(), [None, num_classes])
images = random_ops.random_uniform((batch_size, height, width, 3))
with self.test_session() as sess:
sess.run(variables.global_variables_initializer())
output = sess.run(logits, {inputs: images.eval()})
self.assertEquals(output.shape, (batch_size, num_classes))
def testUnknownImageShape(self):
ops.reset_default_graph()
batch_size = 2
height, width = 224, 224
num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess:
inputs = array_ops.placeholder(
dtypes.float32, shape=(batch_size, None, None, 3))
logits, end_points = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
pre_pool = end_points['Mixed_5c']
feed_dict = {inputs: input_np}
variables.global_variables_initializer().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape), [batch_size, 7, 7, 1024])
def testUnknownImageShape(self):
ops.reset_default_graph()
batch_size = 2
height, width = 224, 224
num_classes = 1000
input_np = np.random.uniform(0, 1, (batch_size, height, width, 3))
with self.test_session() as sess:
inputs = array_ops.placeholder(dtypes.float32,
shape=(batch_size, None, None, 3))
logits, end_points = inception_v2.inception_v2(inputs, num_classes)
self.assertTrue(logits.op.name.startswith('InceptionV2/Logits'))
self.assertListEqual(logits.get_shape().as_list(),
[batch_size, num_classes])
pre_pool = end_points['Mixed_5c']
feed_dict = {inputs: input_np}
variables.global_variables_initializer().run()
pre_pool_out = sess.run(pre_pool, feed_dict=feed_dict)
self.assertListEqual(list(pre_pool_out.shape),
[batch_size, 7, 7, 1024])
