def autoencoder(summary=False):
""" Return the Keras model of the network
Autoencoder reduces the dimensions of data
from (56, 64, 24) to 487
"""
input_shape = (56, 64, 24, 1)
input_img = Input(shape=input_shape)
# x = ZeroPadding3D(padding=((2, 3),(2, 1),(1, 0)))(input_img)
x = Conv3D(16, (3, 3, 3), activation='relu', padding='same')(input_img)
x = MaxPooling3D((2, 2, 2), padding='same')(x)
x = Conv3D(8, (3, 3, 3), activation='relu', padding='same')(x)
x = MaxPooling3D((2, 2, 2), padding='same')(x)
x = Conv3D(8, (3, 3, 3), activation='relu', padding='same')(x)
encoded = MaxPooling3D((2, 2, 2), padding='same')(x)
x = Conv3D(8, (3, 3, 3), activation='relu', padding='same')(encoded)
x = UpSampling3D((2, 2, 2))(x)
x = Conv3D(8, (3, 3, 3), activation='relu', padding='same')(x)
x = UpSampling3D((2, 2, 2))(x)
x = Conv3D(16, (3, 3, 3), activation='relu', padding='same')(x)
x = UpSampling3D((2, 2, 2))(x)
decoded = Conv3D(1, (3, 3, 3), activation='relu', padding='same')(x)
autoencoder = Model(input_img, decoded)
autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')
if summary:
print(autoencoder.summary())
return autoencoder
def conv_3d(self):
"""
Build a 3D convolutional network, based loosely on C3D.
https://arxiv.org/pdf/1412.0767.pdf
"""
model = Sequential()
model.add(
Conv3D(32, (3, 3, 3),
activation='relu',
input_shape=self.input_shape))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Conv3D(64, (3, 3, 3), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Conv3D(128, (3, 3, 3), activation='relu'))
model.add(Conv3D(128, (3, 3, 3), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Conv3D(256, (2, 2, 2), activation='relu'))
model.add(Conv3D(256, (2, 2, 2), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Flatten())
model.add(Dense(1024))
model.add(Dropout(0.5))
# model.add(Dense(1024))
# model.add(Dropout(0.5))
model.add(Dense(self.nb_classes, activation='softmax'))
return model
def build_conv_3d():
model = Sequential()
model.add(
Conv3D(8,
kernel_size=(3, 3, 3),
input_shape=(train_data.shape[1:]),
border_mode='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), border_mode='same'))
model.add(
Conv3D(8,
kernel_size=(3, 3, 3),
input_shape=(train_data.shape[1:]),
border_mode='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), border_mode='same'))
model.add(
Conv3D(8,
kernel_size=(3, 3, 3),
input_shape=(train_data.shape[1:]),
border_mode='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), border_mode='same'))
return model
def cnn(self):
model = Sequential()
samples, frames, x, y, channels = self.input_shape
model.add(
Conv3D(32, (2, 3, 3),
activation='relu',
input_shape=(frames, x, y, channels),
data_format='channels_last'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(2, 2, 2)))
model.add(Conv3D(64, (2, 2, 2), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(2, 3, 3)))
model.add(Conv3D(128, (3, 3, 3), activation='relu'))
model.add(Conv3D(128, (3, 3, 3), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(2, 3, 3)))
model.add(Dropout(0.25))
model.add(Conv3D(256, (2, 2, 2), activation='relu'))
model.add(Conv3D(256, (2, 2, 2), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(2, 3, 3)))
model.add(Flatten())
model.add(Dense(1024))
model.add(Dropout(0.25))
model.add(Dense(1024))
model.add(Dropout(0.25))
model.add(Dense(self.num_classes, activation='softmax'))
return model
def network(input_shape=(49, 49, 49, 1), class_num=2):
"""Example CNN
Keyword Arguments:
input_shape {tuple} -- shape of input images. Should be (28,28,1) for MNIST and (32,32,3) for CIFAR (default: {(28,28,1)})
class_num {int} -- number of classes. Shoule be 10 for both MNIST and CIFAR10 (default: {10})
Returns:
model -- keras.models.Model() object
"""
im_input = Input(shape=input_shape)
t = Convolution3D(32, (24, 24, 24), padding='same')(im_input) # (24,24,24)
t = MaxPooling3D(pool_size=(2, 2, 2))(t)
t = Convolution3D(64, (12, 12, 12), padding='same')(t)
t = MaxPooling3D(pool_size=(2, 2, 2))(t)
t = Convolution3D(128, (6, 6, 6), padding='same')(t)
t = Flatten()(t)
t = Dense(256)(t)
t = Activation('relu')(t)
t = Dense(class_num)(t)
output = Activation('softmax')(t)
model = Model(input=im_input, output=output)
return model
def make_model():
model = Sequential()
#LAYER 1 (convolution 5x5x5)
model.add(
Conv3D(batch_input_shape=[None, 128, 128, 128, 1],
filters=30,
kernel_size=5,
strides=3,
padding='valid',
kernel_regularizer=regularizers.l2(0.01)))
model.add(Activation('relu'))
model.add(MaxPooling3D(pool_size=(2, 2, 2)))
model.add(Dropout(rate=0.1))
model.add(Conv3D(filters=32, kernel_size=5, strides=3, padding='valid'))
model.add(Activation('relu'))
model.add(MaxPooling3D(pool_size=(2, 2, 2)))
model.add(Dropout(rate=0.11))
model.add(Flatten())
model.add(Dense(300, activation='relu'))
model.add(Dropout(rate=0.17))
model.add(Dense(100, activation='relu'))
model.add(Dense(1, kernel_initializer='uniform', activation='sigmoid'))
return model
def create_model(self, verbose=False):
model = Sequential()
model.add(
Conv3D(8, (2, 3, 10),
input_shape=(1, 4, 5, 100),
padding="same",
activation='relu'))
model.add(Conv3D(8, (2, 3, 10), padding="same", activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 1, 5)))
model.add(Conv3D(8, (2, 3, 10), padding="same", activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 1, 2)))
model.add(Dropout(0.3))
model.add(Flatten())
model.add(Dense(32, activation='relu'))
model.add(Dense(16, activation='relu'))
model.add(Dense(4, activation='softmax'))
# Compile model
optimizer = Adam(lr=0.001,
decay=0.001) # lr <- lr * (1.0/(1.0+decay*iter))
model.compile(loss='categorical_crossentropy',
optimizer=optimizer,
metrics=['accuracy'])
if verbose:
model.summary()
return model
def get_model(input_shape=(49, 49, 49, 1), class_num=2):
"""Example CNN
Keyword Arguments:
input_shape {tuple} -- shape of input images. Should be (28,28,1) for MNIST and (32,32,3) for CIFAR (default: {(28,28,1)})
class_num {int} -- number of classes. Shoule be 10 for both MNIST and CIFAR10 (default: {10})
Returns:
model -- keras.models.Model() object
"""
im_input = Input(shape=input_shape)
t = Convolution3D(32, (24, 24, 24), padding='same')(im_input) # (24,24,24)
t = MaxPooling3D(pool_size=(2, 2, 2))(t)
t = Convolution3D(64, (12, 12, 12), padding='same')(t)
t = MaxPooling3D(pool_size=(2, 2, 2))(t)
t = Convolution3D(128, (6, 6, 6), padding='same')(t)
t = Flatten()(t)
t = Dense(256)(t)
t = Activation('relu')(t)
t = Dense(class_num)(t)
output = Activation('softmax')(t)
model = Model(input=im_input, output=output)
sgd = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)
model.compile(sgd, 'categorical_crossentropy', metrics=['accuracy'])
return model
def get_model(self):
model = Sequential()
# 1st layer group
model.add(Conv3D(64, 3, 3, 3, activation='relu',
border_mode='same', name='conv1',
subsample=(1, 1, 1),
input_shape=self.input_shape))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2),
border_mode='valid', name='pool1'))
# 2nd layer group
model.add(Conv3D(128, 3, 3, 3, activation='relu',
border_mode='same', name='conv2',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool2'))
# 3rd layer group
#model.add(Conv3D(256, 3, 3, 3, activation='relu',
# border_mode='same', name='conv3a',
# subsample=(1, 1, 1)))
model.add(Conv3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool3'))
model.add(Flatten())
# FC layers group
model.add(Dense(512, activation='relu', name='fc7'))
model.add(Dropout(0.5))
model.add(Dense(self.output_shape, activation='linear'))
return model
def build(input_shape, num_classes=2):
"""Create a 3D Convolutional Autoencoder model.
Parameters:
- input_shape: Tuple of input shape in the format
(conv_dim1, conv_dim2, conv_dim3, channels)
- initial_filter: Initial filter size. This will be doubled
for each hidden layer as it goes deeper.
- num_encoding_layers: Number of encoding convolutional +
pooling layers. The number of decoding
layers will be the same.
Returns:
- A 3D CAD model that takes a 5D tensor (volumetric images
in batch) as input and returns a 5D vector (prediction) as output.
"""
if len(input_shape) != 4:
raise ValueError("Input shape should be a tuple "
"(conv_dim1, conv_dim2, conv_dim3, channels)")
input_img = Input(shape=input_shape, name="cad_input")
# Conv1 (Output 120 x 120 x 64 x 128)
x = Conv3D(128, (7, 7, 7), activation='relu',
padding='same')(input_img)
x = BatchNormalization()(x)
x = MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2))(x)
# Conv2 (Output 30 x 30 x 16 x 256)
x = Conv3D(256, (5, 5, 5), activation='relu', padding='same')(x)
x = BatchNormalization()(x)
x = MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2))(x)
# Conv3 (Output 7 x 7 x 8 x 256)
x = Conv3D(256, (3, 3, 3), activation='relu', padding='same')(x)
# Conv4 (Output 4 x 4 x 4 x 512)
x = Conv3D(512, (3, 3, 3),
activation='relu',
strides=(2, 2, 2),
padding='same')(x)
# Conv5 (Output 2 x 2 x 2 x 1024)
x = Conv3D(1024, (3, 3, 3),
activation='relu',
strides=(2, 2, 2),
padding='same')(x)
# Pooling and flatten
x = MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2))(x)
x = Flatten()(x)
# Fully connected layers
x = Dense(1024, activation='relu', use_bias=True)(x)
x = Dense(1024, activation='relu', use_bias=True)(x)
x = Dense(num_classes, activation='sigmoid', use_bias=True)(x)
model = Model(inputs=input_img, outputs=x)
return model
def conv_3d(self):
model = Sequential()
model.add(
Conv3D(32, (3, 3, 3),
activation='relu',
input_shape=self.input_shape))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Conv3D(64, (3, 3, 3), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Conv3D(128, (3, 3, 3), activation='relu'))
model.add(Conv3D(128, (3, 3, 3), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Conv3D(256, (2, 2, 2), activation='relu'))
model.add(Conv3D(256, (2, 2, 2), activation='relu'))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
model.add(Flatten())
model.add(Dense(1024))
model.add(Dropout(0.5))
model.add(Dense(1024))
model.add(Dropout(0.5))
model.add(Dense(self.nb_classes, activation='softmax'))
return model
def train_model():
input_img = Input(shape=(128, 128, 128, 1))
x = Convolution3D(16, (5, 5, 5), activation='relu', padding='same')(input_img)
x = MaxPooling3D((2, 2, 2), padding='same')(x)
x = Convolution3D(16, (5, 5, 5), activation='relu', padding='same')(x)
x = MaxPooling3D((2, 2, 2), padding='same')(x)
x = Convolution3D(16, (5, 5, 5), activation='relu', padding='same')(x)
encoded = MaxPooling3D((2, 2, 2), padding='same', name='encoder')(x)
print("shape of encoded: ")
print(K.int_shape(encoded))
x = Convolution3D(16, (5, 5, 5), activation='relu', padding='same')(encoded)
x = UpSampling3D((2, 2, 2))(x)
x = Convolution3D(16, (5, 5, 5), activation='relu', padding='same')(x)
x = UpSampling3D((2, 2, 2))(x)
x = Convolution3D(16, (5, 5, 5), activation='relu', padding='same')(x)
x = UpSampling3D((2, 2, 2))(x)
decoded = Convolution3D(1, (5, 5, 5), activation='sigmoid', padding='same')(x)
print("shape of decoded: ")
print(K.int_shape(decoded))
autoencoder = Model(input_img, decoded)
autoencoder.compile(optimizer='adadelta', loss='binary_crossentropy')
autoencoder.fit(train_data, train_data,
epochs=10,
batch_size=14,
validation_data=(val_data, val_data),
callbacks=[TensorBoard(log_dir='/tmp/autoencoder')])
autoencoder.save('autoencoder.h5')
def createModel(nb_filters=[8, 16, 32, 64, 128, 256],
nb_pool=[2, 3],
nb_conv=[3, 5],
nb_channels=1):
img_depth = len(listOfData[0])
img_cols = len(listOfData[0][0])
img_rows = len(listOfData[0][0][0])
numberOfLabels = len(set(labels))
# CNN Model v3
model = Sequential()
model.add(
Conv3D(nb_filters[1], (nb_conv[0], nb_conv[1], nb_conv[1]),
data_format='channels_last',
input_shape=(img_depth, img_rows, img_cols, nb_channels),
activation='relu'))
model.add(MaxPooling3D(pool_size=(nb_pool[0], nb_pool[1], nb_pool[1])))
model.add(
Conv3D(nb_filters[2], (nb_conv[0], nb_conv[1], nb_conv[1]),
activation='relu'))
model.add(MaxPooling3D(pool_size=(nb_pool[0], nb_pool[1], nb_pool[1])))
model.add(
Conv3D(nb_filters[3], (nb_conv[0], nb_conv[1], nb_conv[1]),
activation='relu'))
model.add(MaxPooling3D(pool_size=(nb_pool[0], nb_pool[1], nb_pool[1])))
model.add(Flatten())
model.add(Dense(128))
model.add(Dropout(0.5))
model.add(Dense(numberOfLabels))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='adadelta',
metrics=['accuracy'])
model.summary()
return model
def cnn_3D(self, input_shape, modual=''):
#建立Sequential模型
model = Sequential()
model.add(Convolution3D(
filters = 8,
kernel_size = (3, 3, 3),
input_shape = input_shape, #40x40x5x1
padding = (0,0,1),
activation='relu',
kernel_initializer='he_normal',
name = modual+'conv1'
))# now 38x38x5x8
model.add(MaxPooling3D(pool_size=(2,2,1)))# now 19x19x5x8
model.add(Convolution3D(
filters = 16,
kernel_size = (4, 4, 3),
activation='relu',
kernel_initializer='he_normal',
name = modual+'conv2'
))# now 16x16x3x16
model.add(MaxPooling3D(pool_size=(2,2,1)))# now 8x8x1x16
model.add(Convolution3D(
filters = 32,
kernel_size = (3, 3, 3),
activation='relu',
kernel_initializer='he_normal',
name = modual+'conv2'
))# now 6x6x1x32
model.add(AveragePooling3D(pool_size=(6,6,1)))# now 1x1x1x32
model.add(Flatten())
# model.add(Dropout(0.5))
model.add(Dense(32, activation='relu', name = modual+'fc1'))
return model
def c3d(self):
# Tunable parameters
strides = (1, 1, 1)
model = Sequential()
model.add(Conv3D(32, kernel_size=(3, 3, 3), strides=strides,input_shape=(
self.input_shape), border_mode='same', activation='relu'))
#model.add(Activation('relu'))
model.add(Conv3D(32, kernel_size=(3, 3, 3), strides=strides,padding='same', activation='softmax'))
#model.add(Activation('softmax'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2), padding='same'))
model.add(Dropout(0.25))
model.add(Conv3D(64, kernel_size=(3, 3, 3),strides=strides, padding='same', activation='relu'))
#model.add(Activation('relu'))
model.add(Conv3D(64, kernel_size=(3, 3, 3),strides=strides, padding='same', activation='softmax'))
#model.add(Activation('softmax'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2), padding='same'))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(512, activation='sigmoid'))
model.add(Dropout(0.5))
model.add(Dense(self.n_labels, activation='softmax'))
return model
def get_model(self):
model = Sequential()
model.add(
Conv3D(64,
kernel_size=(3, 3, 3),
input_shape=self.input_shape,
activation='relu'))
model.add(MaxPooling3D(pool_size=2, strides=(1, 2, 2)))
model.add(
Conv3D(64,
kernel_size=(3, 2, 2),
input_shape=self.input_shape,
activation='relu'))
model.add(Conv3D(128, kernel_size=(3, 2, 2), activation='relu'))
model.add(MaxPooling3D(pool_size=2, strides=(1, 2, 2)))
model.add(Conv3D(128, kernel_size=(3, 2, 2), activation='relu'))
model.add(MaxPooling3D(pool_size=2, strides=2))
model.add(Dropout(0.15))
model.add(Conv3D(256, kernel_size=(3, 2, 2), activation='relu'))
model.add(BatchNormalization())
model.add(Flatten())
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.15))
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.15))
model.add(Dense(self.output_shape, activation='linear'))
return model
def construct_model(self):
# th input (channels, height, width)
# tf input (height, width, channels)
keras.backend.set_image_dim_ordering('tf')
model = Sequential()
#Encode:
model.add(Convolution3D(16,(2,3,3), activation='relu', input_shape=(config.frames_per_input, 501, 501, 3),border_mode='same'))
model.add(MaxPooling3D(pool_size=(2,2,2)))
model.add(Convolution3D(16,(2,3,3), activation='relu', border_mode='same'))
model.add(Convolution3D(16,(2,3,3), activation='relu', border_mode='same'))
model.add(MaxPooling3D(pool_size=(2,2,2)))
model.add(Convolution3D(8,(1,3,3), activation='relu', border_mode='same'))
#model.add(Convolution3D(8,(1,3,3), activation='relu', border_mode='same'))
#model.add(MaxPooling3D(pool_size=(1,2,2)))
#model.add(Convolution3D(8,(1,3,3), activation='relu', border_mode='same'))
#model.add(Convolution3D(8,(1,3,3), activation='relu', border_mode='same'))
#Decode:
model.add(Convolution3D(16,(2,3,3), activation='relu', border_mode='same'))
model.add(UpSampling3D((1,2,2)))
model.add(Convolution3D(16,(2,3,3), activation='relu', border_mode='same'))
model.add(UpSampling3D((1,2,2)))
model.add(Convolution3D(3,(2,3,3), activation='relu', border_mode='same'))
# model.add(Convolution3D(16,(2,3,3), activation='relu', border_mode='same'))
# model.add(UpSampling3D((1,1,1)))
# model.add(Convolution3D(3,(2,3,3), activation='relu', border_mode='same'))
# model.add(UpSampling3D((1,1,1)))
# model.add(Flatten())
model.compile(loss='mean_squared_error', optimizer='adam')
model.compile(loss=y_std, optimizer='adam')
model.compile(loss='binary_crossentropy', optimizer='adam') # doesn't reset weights
model.summary()
self.model = model
def make_model(l2_penalty):
model = Sequential()
#LAYER 1 (convolution 5x5x5)
model.add(
Conv3D(batch_input_shape=[None, 128, 128, 128, 1],
filters=16,
kernel_size=5,
strides=3,
padding='valid',
kernel_regularizer=l2(l2_penalty),
activation='relu'))
model.add(MaxPooling3D(pool_size=(2, 2, 2)))
model.add(
Conv3D(filters=32,
kernel_size=5,
strides=3,
padding='valid',
activation='relu'))
model.add(MaxPooling3D(pool_size=(2, 2, 2)))
model.add(Flatten())
model.add(Dense(100, activation='relu'))
model.add(Dropout(rate=0.15))
model.add(Dense(100, activation='relu'))
model.add(Dense(1, kernel_initializer='uniform', activation='sigmoid'))
print(model.output_shape)
return model
def c3d_model(summary=False):
""" Return the Keras model of the network
"""
model = Sequential()
# 1st layer group
model.add(Convolution3D(64, 3, 3, 3, activation='relu',
border_mode='same', name='conv1',
subsample=(1, 1, 1),
input_shape=(16,120,120,3)))
# model.add(BatchNormalization())
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2),
border_mode='same', name='pool1'))
# 2nd layer group
model.add(Convolution3D(128, 3, 3, 3, activation='relu',
border_mode='same', name='conv2',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='same', name='pool2'))
# 3rd layer group
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3a',
subsample=(1, 1, 1)))
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='same', name='pool3'))
# 4th layer group
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4a',
subsample=(1, 1, 1)))
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='same', name='pool4'))
# 5th layer group
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5a',
subsample=(1, 1, 1)))
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5b',
subsample=(1, 1, 1)))
model.add(ZeroPadding3D(padding=(0, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool5'))
model.add(BatchNormalization())
model.add(Flatten())
# FC layers group
model.add(Dense(4096,activation='relu', name='fc6'))
model.add(Dropout(.5))
model.add(BatchNormalization())
model.add(Dense(4096, activation='relu', name='fc7'))
model.add(Dropout(.5))
model.add(Dense(14, activation='softmax', name='fc9'))
if summary:
print(model.summary())
return model
def Conv3dBLSTM(self) :
model = Sequential()
# 1st layer group
model = Sequential()
# 1st layer group
# (samples, conv_dim1, conv_dim2, conv_dim3, channels)
model.add(Conv3D(64, 3, 3, 3, activation='relu',
border_mode='same', name='conv1',
subsample=(1, 1, 1),
input_shape=self.shapeOfInputConv3d)) # 40 80 80 3
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2),
border_mode='valid', name='pool1'))
# 2nd layer group
model.add(Conv3D(128, 3, 3, 3, activation='relu',
border_mode='same', name='conv2',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool2'))
# 3rd layer group
model.add(Conv3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3a',
subsample=(1, 1, 1)))
model.add(Conv3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool3'))
# 4th layer group
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4a',
subsample=(1, 1, 1)))
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool4'))
# 5th layer group
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5a',
subsample=(1, 1, 1)))
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5b',
subsample=(1, 1, 1)))
model.add(ZeroPadding3D(padding=(0, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool5'))
model.add(Reshape(target_shape=(2, 512* 3* 3)))
# (batch_size, timesteps, input_dim)
model.add(Bidirectional(LSTM(512* 3* 3, input_shape=self.shapeOfInputLSTM,
return_sequences=False,
activation="softmax",
dropout=0.5)))
model.add(Dense(512, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(self.classesNumber, activation='softmax'))
return model
def build(input_shape, num_outputs):
print('original input shape:', input_shape)
_handle_dim_ordering()
if len(input_shape) != 4:
raise Exception(
"Input shape should be a tuple (nb_channels, kernel_dim1, kernel_dim2, kernel_dim3)"
)
print('original input shape:', input_shape)
# orignal input shape: 1,7,7,200
if K.image_dim_ordering() == 'tf':
input_shape = (input_shape[1], input_shape[2], input_shape[3],
input_shape[0])
print('change input shape:', input_shape)
# 用keras中函数式模型API，不用序贯模型API
# 张量流输入
input = Input(shape=input_shape)
conv1 = Conv3D(filters=32,
kernel_size=(3, 3, 20),
strides=(1, 1, 5),
kernel_regularizer=regularizers.l2(0.01))(input)
act1 = Activation('relu')(conv1)
pool1 = MaxPooling3D(pool_size=(2, 2, 2),
strides=(1, 1, 1),
padding='same')(act1)
conv2 = Conv3D(filters=64,
kernel_size=(2, 2, 3),
strides=(1, 1, 2),
kernel_regularizer=regularizers.l2(0.01))(pool1)
act2 = Activation('relu')(conv2)
drop1 = Dropout(0.5)(act2)
pool2 = MaxPooling3D(pool_size=(2, 2, 2),
strides=(1, 1, 1),
padding='same')(drop1)
conv3 = Conv3D(filters=128,
kernel_size=(3, 3, 3),
strides=(1, 1, 2),
kernel_regularizer=regularizers.l2(0.01))(pool2)
act3 = Activation('relu')(conv3)
drop2 = Dropout(0.5)(act3)
flatten1 = Flatten()(drop2)
fc1 = Dense(200, kernel_regularizer=regularizers.l2(0.01))(flatten1)
act3 = Activation('relu')(fc1)
# 输入分类器
# Classifier block
dense = Dense(units=num_outputs,
activation="softmax",
kernel_initializer="he_normal")(act3)
model = Model(inputs=input, outputs=dense)
return model
def C3DP_4_DE_3_V1(n_labels, input_shape):
#image size (26,39) and seq length 32
# Tunable parameters
model_name = 'C3DP_4_DE_3_V1'
kernel_size = (3, 3, 3)
strides = (1, 1, 1)
extra_conv_blocks = 1
model = Sequential()
# 3D Conv Block 1 + Max Pooling
model.add(
Conv3D(32, (7, 7, 5),
strides=strides,
activation='relu',
padding='same',
input_shape=input_shape))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
# 3D Conv Block 2 + Max Pooling
model.add(Conv3D(64, (5, 5, 3),
strides=strides,
activation='relu',
padding='same'),
input_shape=(32, 13, 19, 3))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2)))
# 3D Conv Block 3 + Max Pooling
model.add(
Conv3D(128, (5, 5, 3),
strides=strides,
activation='relu',
padding='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2)))
# 3D Conv Block 4 + Max Pooling + Dropout
model.add(
Conv3D(128, (3, 5, 3),
strides=strides,
activation='relu',
padding='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
# Dense Block 1
model.add(Dense(256, activation='relu'))
# Dense Block 2 + Dropout
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.2))
# Dense Block 3
model.add(Dense(n_labels, activation='softmax'))
return (model, model_name)
def C3D_2_C3DP_2_DE_2(n_labels, input_shape):
#image size (26,39) and seq length 32
# Tunable parameters
model_name = 'C3D_2_C3DP_2_DE_2'
strides = (1, 1, 1)
model = Sequential()
# 3D Conv Block 1
model.add(
Conv3D(32,
kernel_size=(3, 3, 3),
strides=strides,
input_shape=input_shape,
border_mode='same',
activation='relu'))
# 3D Conv Block 2 + Max Pooling + Dropout
model.add(
Conv3D(32,
kernel_size=(3, 3, 3),
strides=strides,
padding='same',
activation='softmax'))
model.add(
MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2), padding='same'))
model.add(Dropout(0.25))
# 3D Conv Block 3
model.add(
Conv3D(64,
kernel_size=(3, 3, 3),
strides=strides,
padding='same',
activation='relu'))
# 3D Conv Block 4 + Max Pooling + Dropout
model.add(
Conv3D(64,
kernel_size=(3, 3, 3),
strides=strides,
padding='same',
activation='softmax'))
model.add(
MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2), padding='same'))
model.add(Dropout(0.25))
model.add(Flatten())
#Dense Block 1 + Dropout
model.add(Dense(512, activation='sigmoid'))
model.add(Dropout(0.5))
#Dense Block 2
model.add(Dense(n_labels, activation='softmax'))
return (model, model_name)
def C3D(weights='sports1M'):
"""Instantiates a C3D Kerasl model
Keyword arguments:
weights -- weights to load into model. (default is sports1M)
Returns:
A Keras model.
"""
if weights not in {'sports1M', None}:
raise ValueError('weights should be either be sports1M or None')
if K.image_data_format() == 'channels_last':
shape = (16,112,112,3)
else:
shape = (3,16,112,112)
model = Sequential()
model.add(Conv3D(64, 3, activation='relu', padding='same', name='conv1', input_shape=shape))
model.add(MaxPooling3D(pool_size=(1,2,2), strides=(1,2,2), padding='same', name='pool1'))
model.add(Conv3D(128, 3, activation='relu', padding='same', name='conv2'))
model.add(MaxPooling3D(pool_size=(2,2,2), strides=(2,2,2), padding='valid', name='pool2'))
model.add(Conv3D(256, 3, activation='relu', padding='same', name='conv3a'))
model.add(Conv3D(256, 3, activation='relu', padding='same', name='conv3b'))
model.add(MaxPooling3D(pool_size=(2,2,2), strides=(2,2,2), padding='valid', name='pool3'))
model.add(Conv3D(512, 3, activation='relu', padding='same', name='conv4a'))
model.add(Conv3D(512, 3, activation='relu', padding='same', name='conv4b'))
model.add(MaxPooling3D(pool_size=(2,2,2), strides=(2,2,2), padding='valid', name='pool4'))
model.add(Conv3D(512, 3, activation='relu', padding='same', name='conv5a'))
model.add(Conv3D(512, 3, activation='relu', padding='same', name='conv5b'))
model.add(ZeroPadding3D(padding=(0,1,1)))
model.add(MaxPooling3D(pool_size=(2,2,2), strides=(2,2,2), padding='valid', name='pool5'))
model.add(Flatten())
model.add(Dense(4096, activation='relu', name='fc6'))
model.add(Dropout(0.5))
model.add(Dense(4096, activation='relu', name='fc7'))
model.add(Dropout(0.5))
model.add(Dense(487, activation='softmax', name='fc8'))
if weights == 'sports1M':
weights_path = get_file('sports1M_weights_tf.h5',
WEIGHTS_PATH,
cache_subdir='models',
md5_hash='b7a93b2f9156ccbebe3ca24b41fc5402')
model.load_weights(weights_path)
return model
def c3d_model():
classes = 2
""" Return the Keras model of the network
"""
model = Sequential()
input_shape=(16, 100, 100, 3) # l, h, w, c
model.add(Convolution3D(64, 3, 3, 3, activation='relu',
border_mode='same', name='conv1',
input_shape=input_shape))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2),
border_mode='valid', name='pool1'))
# 2nd layer group
model.add(Convolution3D(128, 3, 3, 3, activation='relu',
border_mode='same', name='conv2'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool2'))
# 3rd layer group
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3a'))
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3b'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool3'))
# 4th layer group
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv4a'))
model.add(Convolution3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv4b'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool4'))
# 5th layer group
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5a'))
model.add(Convolution3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5b'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool5'))
model.add(Flatten())
# FC layers group
#model.add(Dense(2048, activation='relu', name='fc6'))
#model.add(Dropout(.5))
#model.add(Dense(2048, activation='relu', name='fc7'))
#model.add(Dropout(.5))
#model.add(Dense(2, activation='sigmoid', name='fc8'))
#WS FC Layer with custom parameters
model.add(Dense(512,input_dim=4608,W_regularizer=l2(0.001),activation='relu', name='fc6'))
model.add(Dropout(.6))
model.add(Dense(32,activation='relu',W_regularizer=l2(0.001), name='fc7'))
model.add(Dropout(.6))
model.add(Dense(2, activation='sigmoid',W_regularizer=l2(0.001), name='fc8'))
return model
def small_c3d(self):
"""See: 'https://arxiv.org/pdf/1412.0767.pdf' """
# Tunable parameters
kernel_size = (3, 3, 3)
strides = (1, 1, 1)
extra_conv_blocks = 1
model = Sequential()
# Conv Block 1
model.add(
Conv3D(32,
kernel_size,
strides=strides,
activation='relu',
padding='same',
input_shape=self.input_shape))
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2)))
# Conv Block 2
model.add(
Conv3D(64,
kernel_size,
strides=strides,
activation='relu',
padding='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2)))
# Conv Block 3
model.add(
Conv3D(128,
kernel_size,
strides=strides,
activation='relu',
padding='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2)))
# Conv Block 4
model.add(
Conv3D(128,
kernel_size,
strides=strides,
activation='relu',
padding='same'))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2)))
# Dense Block
model.add(Flatten())
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(256, activation='relu'))
model.add(Dropout(0.25))
model.add(Dense(self.n_labels, activation='softmax'))
return model
def Conv3DModelCreate(self):
# According to https://gist.github.com/albertomontesg/d8b21a179c1e6cca0480ebdf292c34d2
model = Sequential()
# 1st layer group
model.add(Conv3D(64, 3, 3, 3, activation='relu',
border_mode='same', name='conv1',
subsample=(1, 1, 1),
input_shape=self.shapeOfInput)) # 40 80 80 3
model.add(MaxPooling3D(pool_size=(1, 2, 2), strides=(1, 2, 2),
border_mode='valid', name='pool1'))
# 2nd layer group
model.add(Conv3D(128, 3, 3, 3, activation='relu',
border_mode='same', name='conv2',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool2'))
# 3rd layer group
model.add(Conv3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3a',
subsample=(1, 1, 1)))
model.add(Conv3D(256, 3, 3, 3, activation='relu',
border_mode='same', name='conv3b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool3'))
# 4th layer group
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4a',
subsample=(1, 1, 1)))
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv4b',
subsample=(1, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool4'))
# 5th layer group
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5a',
subsample=(1, 1, 1)))
model.add(Conv3D(512, 3, 3, 3, activation='relu',
border_mode='same', name='conv5b',
subsample=(1, 1, 1)))
model.add(ZeroPadding3D(padding=(0, 1, 1)))
model.add(MaxPooling3D(pool_size=(2, 2, 2), strides=(2, 2, 2),
border_mode='valid', name='pool5'))
model.add(Flatten())
# FC layers group
model.add(Dense(4096, activation='relu', name='fc6'))
model.add(Dropout(.5))
model.add(Dense(4096, activation='relu', name='fc7'))
model.add(Dropout(.5))
model.add(Dense(self.classesNumber, activation='softmax', name='fc8'))
return model
def build_3d_model(nb_classes):
#kernal_size = (depth,row,col)
kernal_size = (4,3,3)
pool_size = (1,2,2)
timesteps =10
#input_shape = (timesteps,nb_sensors,rows,cols)
input_shape = (timesteps,4,9,5)
model = Sequential()
#1st layer group
model.add(Conv3D(64,kernal_size, name = 'conv1',
activation = 'relu', input_shape = input_shape,
strides = (1,1,1), data_format = 'channels_first',
padding = 'same'))
model.add(MaxPooling3D(padding = 'valid',pool_size=pool_size,
strides = (1,2,2),data_format = 'channels_first',name='pool1'))
model.add(Dropout(0.25))
#2nd layer group
model.add(Conv3D(128,kernal_size,strides = (1,1,1),
data_format = 'channels_first',input_shape=input_shape,
activation='relu', padding = 'same', name = 'conv2'))
model.add(MaxPooling3D(pool_size=pool_size,strides = (1,2,2),
padding = 'valid',data_format = 'channels_first',name='pool2'))
model.add(Dropout(0.25))
#3rd layer group
model.add(Conv3D(256,kernal_size,strides = (1,1,1),
data_format = 'channels_first',input_shape=input_shape,
activation='relu', padding = 'same', name = 'conv3'))
#model.add(MaxPooling3D(pool_size=pool_size,strides = (1,2,2),
# padding = 'same',name='pool3'))
model.add(Dropout(0.25))
#FC layers
model.add(Flatten())
model.add(Dense(256,activation='relu',name='fc1'))
model.add(Dropout(0.5))
model.add(Dense(nb_classes,activation='relu',name='fc2'))
model.add(Dropout(0.5))
print(model.summary())
start = time.time()
model.compile(loss='categorical_crossentropy',
optimizer='adadelta',
metrics=['accuracy'])
print("> Compile time :",time.time()-start)
return model
def c3d(self):
""" Return the Keras model of the network
"""
model = Sequential()
# 1st layer group
model.add(
Conv3D(64, (3, 3, 3),
activation='relu',
padding='same',
name='3d1',
strides=(1, 1, 1),
input_shape=self.input_shape))
model.add(
MaxPooling3D(pool_size=(1, 2, 2),
strides=(1, 2, 2),
padding='valid',
name='pool1'))
# 2nd layer group
model.add(
Conv3D(128, (3, 3, 3),
activation='relu',
padding='same',
name='3d2',
strides=(1, 1, 1)))
model.add(
MaxPooling3D(pool_size=(2, 2, 2),
strides=(2, 2, 2),
padding='valid',
name='pool2'))
# 3rd layer group
model.add(
Conv3D(256, (3, 3, 3),
activation='relu',
padding='same',
name='3d3',
strides=(2, 2, 2)))
model.add(
MaxPooling3D(pool_size=(4, 4, 4),
strides=(2, 2, 2),
padding='valid',
name='pool3'))
model.add(Flatten())
# FC layers group
model.add(Dense((2048), activation='relu', name='fc1'))
model.add(Dropout(0.2))
model.add(Dense((1024), activation='relu', name='fc2'))
model.add(Dropout(0.2))
model.add(Dense((512), activation='relu', name='fc3'))
model.add(Dropout(0.2))
model.add(Dense(self.classes, activation='softmax'))
return model
def get_model(self):
invariants = {'kernel_initializer': 'he_normal'}
#model = Sequential()
seq = Input(shape=self.input_shape)
grey = TimeDistributed(
Lambda(lambda x: tf.image.rgb_to_grayscale(x)))(seq)
red = Lambda(lambda x: x[:, :, :, :, 0])(seq)
red = Lambda(lambda x: x[:, :, :, :, None])(red)
x1 = Conv3D(64, kernel_size=(15, 5, 5), padding='valid',
**invariants)(grey)
x1 = Activation('relu')(x1)
x1 = MaxPooling3D(pool_size=(1, 2, 2),
strides=(1, 2, 2),
padding='valid')(x1)
x1 = BatchNormalization()(x1)
x1 = Conv3D(128, kernel_size=5, padding='valid', **invariants)(x1)
x1 = Activation('relu')(x1)
x1 = MaxPooling3D(pool_size=(1, 2, 2),
strides=(1, 2, 2),
padding='valid')(x1)
x1 = BatchNormalization()(x1)
x2 = Conv3D(64, kernel_size=(15, 5, 5), padding='valid',
**invariants)(red)
x2 = Activation('relu')(x2)
x2 = MaxPooling3D(pool_size=(1, 2, 2),
strides=(1, 2, 2),
padding='valid')(x2)
x2 = BatchNormalization()(x2)
x2 = Conv3D(128, kernel_size=5, padding='valid', **invariants)(x2)
x2 = Activation('relu')(x2)
x2 = MaxPooling3D(pool_size=(1, 2, 2),
strides=(1, 2, 2),
padding='valid')(x2)
x2 = BatchNormalization()(x2)
x = concatenate([x1, x2])
x = Conv3D(256, kernel_size=3, padding='valid', **invariants)(x)
x = Activation('relu')(x)
x = MaxPooling3D(pool_size=2, strides=(1, 2, 2), padding='valid')(x)
x = BatchNormalization()(x)
x = Flatten()(x)
x = Dense(512, activation='relu', **invariants)(x)
x = Dropout(0.15)(x)
x = Dense(512, activation='relu', **invariants)(x)
x = Dropout(0.15)(x)
pred = Dense(self.output_shape, activation='linear')(x)
model = Model(inputs=seq, outputs=pred)
return model
