def cnn_model(logits=False, input_ph=None, img_rows=28, img_cols=28,
channels=1, nb_filters=64, nb_classes=10):
"""
Defines a CNN model using Keras sequential model
:param logits: If set to False, returns a Keras model, otherwise will also
return logits tensor
:param input_ph: The TensorFlow tensor for the input
(needed if returning logits)
("ph" stands for placeholder but it need not actually be a
placeholder)
:param img_rows: number of row in the image
:param img_cols: number of columns in the image
:param channels: number of color channels (e.g., 1 for MNIST)
:param nb_filters: number of convolutional filters per layer
:param nb_classes: the number of output classes
:return:
"""
model = Sequential()
# Define the layers successively (convolution layers are version dependent)
if keras.backend.image_dim_ordering() == 'th':
input_shape = (channels, img_rows, img_cols)
else:
input_shape = (img_rows, img_cols, channels)
layers = [conv_2d(nb_filters, (5, 5), (1, 1), "same",
input_shape=input_shape),
Activation('relu'),
conv_2d(nb_filters, (5, 5), (1, 1), "valid"),
Activation('relu'),
Flatten(),
Dropout(0.25),
Dense(128),
Activation('relu'),
Dropout(0.5),
Dense(nb_classes)]
for layer in layers:
model.add(layer)
if logits:
logits_tensor = model(input_ph)
model.add(Activation('softmax'))
if logits:
return model, logits_tensor
else:
return model
def cleverhans_model(input_shape, nb_filters, nb_classes, logits,
input_range_type, pre_filter):
"""
Defines a CNN model using Keras sequential model
:params logits: return logits(input of softmax layer) if True; return softmax output otherwise.
:params input_range_type: the expected input range, {1: [0,1], 2:[-0.5, 0.5], 3:[-1, 1]...}
:return:
"""
model = Sequential()
if input_range_type == 1:
# The input data range is [0, 1].
# Don't need to do scaling for cleverhans models, as it is the input range by default.
scaler = lambda x: x
elif input_range_type == 2:
# The input data range is [-0.5, 0.5]. Convert to [0,1] by adding 0.5 element-wise.
scaler = lambda x: x + 0.5
elif input_range_type == 3:
# The input data range is [-1, 1]. Convert to [0,1] by x/2+0.5.
scaler = lambda x: x / 2 + 0.5
layers = [Lambda(scaler, input_shape=input_shape)]
layers += [Lambda(pre_filter, output_shape=input_shape)]
layers += [
Dropout(0.2),
conv_2d(nb_filters, (8, 8), (2, 2), "same"),
Activation('relu'),
conv_2d((nb_filters * 2), (6, 6), (2, 2), "valid"),
Activation('relu'),
conv_2d((nb_filters * 2), (5, 5), (1, 1), "valid"),
Activation('relu'),
Dropout(0.5),
Flatten(),
Dense(nb_classes)
]
for layer in layers:
model.add(layer)
if not logits:
model.add(Activation('softmax'))
return model
def conv_2d(*args, **kwargs):
from cleverhans.utils_keras import conv_2d
warnings.warn("utils.conv_2d is deprecated and may be removed on or after"
" 2018-01-05. Switch to utils_keras.conv_2d.")
return conv_2d(*args, **kwargs)