def testModelHasExpectedNumberOfParameters(self):
batch_size = 5
height, width = 224, 224
inputs = tf.random_uniform((batch_size, height, width, 3))
with slim.arg_scope(inception_v1.inception_v1_arg_scope()):
inception_v1.inception_v1_base(inputs)
total_params, _ = slim.model_analyzer.analyze_vars(
slim.get_model_variables())
self.assertAlmostEqual(5607184, total_params)
def testBuildAndCheckAllEndPointsUptoMixed5c(self):
batch_size = 5
height, width = 224, 224
inputs = tf.random_uniform((batch_size, height, width, 3))
_, end_points = inception_v1.inception_v1_base(
inputs, final_endpoint='Mixed_5c')
endpoints_shapes = {
'Conv2d_1a_7x7': [5, 112, 112, 64],
'MaxPool_2a_3x3': [5, 56, 56, 64],
'Conv2d_2b_1x1': [5, 56, 56, 64],
'Conv2d_2c_3x3': [5, 56, 56, 192],
'MaxPool_3a_3x3': [5, 28, 28, 192],
'Mixed_3b': [5, 28, 28, 256],
'Mixed_3c': [5, 28, 28, 480],
'MaxPool_4a_3x3': [5, 14, 14, 480],
'Mixed_4b': [5, 14, 14, 512],
'Mixed_4c': [5, 14, 14, 512],
'Mixed_4d': [5, 14, 14, 512],
'Mixed_4e': [5, 14, 14, 528],
'Mixed_4f': [5, 14, 14, 832],
'MaxPool_5a_2x2': [5, 7, 7, 832],
'Mixed_5b': [5, 7, 7, 832],
'Mixed_5c': [5, 7, 7, 1024]
}
self.assertItemsEqual(endpoints_shapes.keys(), end_points.keys())
for endpoint_name in endpoints_shapes:
expected_shape = endpoints_shapes[endpoint_name]
self.assertTrue(endpoint_name in end_points)
self.assertListEqual(
end_points[endpoint_name].get_shape().as_list(),
expected_shape)
def __init__(self, inputs, labels, num_frame):
# inputs is a placeholder shape=[batch_size, None, 640, 380, 3]
self.input = inputs
self.feature_maps = []
self.label = labels
self.label = labels
# load the parameters in the session
# calculate the convs
# self.net, self.endpoints = inception_v1.inception_v1_base(inputs=self.input,
# final_endpoint='Mixed_5c',
# scope='InceptionV1')
# prepare the feature map of videos
for i in range(self.input.shape[0]):
tmp = []
for j in range(num_frame[i]):
net = inception_v1.inception_v1_base(inputs=self.input[i][j],
final_endpoint='Mixed_5c',
scope='InceptionV1')
net = tf.reshape(net, shape=[1, -1])
tmp.append(net)
self.feature_maps.append(tmp)
self._rnn_ = self.get_dynamic_lstm()
self._initial_state_ = self._rnn_.zero_state(
rnn_batch_size, LSTMAutoencoder.data_type())
self.state = self._initial_state_
self._output, self.state = tf.nn.dynamic_rnn(
self._rnn_,
inputs=self.feature_maps,
dtype=LSTMAutoencoder.data_type())
# self.output's shape is [batch_size, max_video_length, 1024]
# then we must get each last cell's output of the videos
self.lstm_output = []
for i in range(self.input.shape[0]):
j = self.input[i].shape[0]
self.lstm_output.append(self._output[i][j - 1])
weight = tf.Variable(
tf.truncated_normal([1024, 9],
mean=0.1,
stddev=1.0,
dtype=tf.float32))
bias = tf.Variable(
tf.truncated_normal(shape=[10],
mean=0.1,
stddev=1.0,
dtype=tf.float32))
y_ = []
for _ in self.input.shape[0]:
y_.append(tf.matmul(self.lstm_output[_], weight) + bias)
self.loss = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(y_, self.label))
self.train = tf.train.GradientDescentOptimizer(0.9).minimize(self.loss)
def testHalfSizeImages(self):
batch_size = 5
height, width = 112, 112
inputs = tf.random_uniform((batch_size, height, width, 3))
mixed_5c, _ = inception_v1.inception_v1_base(inputs)
self.assertTrue(mixed_5c.op.name.startswith('InceptionV1/Mixed_5c'))
self.assertListEqual(mixed_5c.get_shape().as_list(),
[batch_size, 4, 4, 1024])
def testBuildBaseNetwork(self):
batch_size = 5
height, width = 224, 224
inputs = tf.random_uniform((batch_size, height, width, 3))
mixed_6c, end_points = inception_v1.inception_v1_base(inputs)
self.assertTrue(mixed_6c.op.name.startswith('InceptionV1/Mixed_5c'))
self.assertListEqual(mixed_6c.get_shape().as_list(),
[batch_size, 7, 7, 1024])
expected_endpoints = [
'Conv2d_1a_7x7', 'MaxPool_2a_3x3', 'Conv2d_2b_1x1',
'Conv2d_2c_3x3', 'MaxPool_3a_3x3', 'Mixed_3b', 'Mixed_3c',
'MaxPool_4a_3x3', 'Mixed_4b', 'Mixed_4c', 'Mixed_4d', 'Mixed_4e',
'Mixed_4f', 'MaxPool_5a_2x2', 'Mixed_5b', 'Mixed_5c'
]
self.assertItemsEqual(end_points.keys(), expected_endpoints)
def testBuildOnlyUptoFinalEndpoint(self):
batch_size = 5
height, width = 224, 224
endpoints = [
'Conv2d_1a_7x7', 'MaxPool_2a_3x3', 'Conv2d_2b_1x1',
'Conv2d_2c_3x3', 'MaxPool_3a_3x3', 'Mixed_3b', 'Mixed_3c',
'MaxPool_4a_3x3', 'Mixed_4b', 'Mixed_4c', 'Mixed_4d', 'Mixed_4e',
'Mixed_4f', 'MaxPool_5a_2x2', 'Mixed_5b', 'Mixed_5c'
]
for index, endpoint in enumerate(endpoints):
with tf.Graph().as_default():
inputs = tf.random_uniform((batch_size, height, width, 3))
out_tensor, end_points = inception_v1.inception_v1_base(
inputs, final_endpoint=endpoint)
self.assertTrue(
out_tensor.op.name.startswith('InceptionV1/' + endpoint))
self.assertItemsEqual(endpoints[:index + 1], end_points.keys())
