def __init__(self,**kwargs):
self.kwargs = kwargs
self.name = kwargs['name']
self.filters = kwargs['filters']
self.kernel = kwargs['kernel_size']
self.activation = 'linear'
if 'activation' in kwargs:
self.activation = kwargs['activation']
kwargs['activation'] = 'linear'
if isinstance(self.kernel, int):
ks = self.kernel
self.kernel = (ks, ks, ks)
ks = self.kernel
kwargs['kernel_size'] = (1, ks[1], ks[2])
kwargs['name'] = self.name + '_x_axis'
kwargs['data_format'] = 'channels_first'
self.convx = KL.Convolution3D(**kwargs)
kwargs['kernel_size'] = (ks[0], 1, ks[2])
kwargs['name'] = self.name + '_y_axis'
kwargs['data_format'] = 'channels_first'
self.convy = KL.Convolution3D(**kwargs)
kwargs['kernel_size'] = (ks[0], ks[1], 1)
kwargs['name'] = self.name + '_z_axis'
kwargs['data_format'] = 'channels_first'
self.convz = KL.Convolution3D(**kwargs)
self.addLayer = KL.Add(name=self.name + '_add')
def _build_model(self):
inputs = layers.Input(shape=(3, self.sequence_size, self.img_size, self.img_size))
x = layers.Convolution3D(16, 3, 3, 3, activation=Relu, border_mode="same")(inputs)
x = layers.Convolution3D(1, 3, 3, 3, activation=Relu, border_mode="same")(x)
x = layers.Reshape([self.sequence_size,321,321])(x)
x = layers.Convolution2D(64, 3, 3, activation=Relu, border_mode="same")(x)
x = layers.MaxPooling2D((2, 2))(x)
x = layers.Convolution2D(32, 3, 3, activation=Relu, border_mode="same")(x)
x = layers.Convolution2D(32, 3, 3, activation=Relu)(x)
x = layers.Convolution2D(4, 3, 3, activation=Relu, border_mode="same")(x)
x = layers.Convolution2D(1, 3, 3, activation='sigmoid', border_mode="same")(x)
model = Model(input=inputs, output=x)
return model
def make_model():
model = models.Sequential()
model.add(
layers.Convolution3D(16,
1,
3,
3,
input_shape=(1, 24, 128, 128),
activation='relu'))
model.add(layers.Convolution3D(32, 1, 3, 3, activation='relu'))
model.add(layers.MaxPooling3D(pool_size=(1, 2, 2)))
model.add(layers.Convolution3D(32, 1, 3, 3, activation='relu'))
model.add(layers.Convolution3D(64, 1, 3, 3, activation='relu'))
model.add(layers.MaxPooling3D(pool_size=(1, 2, 2)))
model.add(layers.Convolution3D(64, 1, 3, 3, activation='relu'))
model.add(layers.Convolution3D(128, 1, 3, 3, activation='relu'))
model.add(layers.MaxPooling3D(pool_size=(1, 2, 2)))
model.add(layers.Convolution3D(128, 1, 3, 3, activation='relu'))
model.add(layers.MaxPooling3D(pool_size=(1, 2, 2)))
model.add(layers.Flatten())
model.add(layers.Dense(512, activation='relu'))
model.add(layers.Dropout(0.5))
model.add(layers.Dense(2, activation='softmax'))
model.compile(loss='categorical_crossentropy',
optimizer='adadelta',
metrics=['accuracy'])
model.fit(trn_x,
trn_y,
batch_size=24,
nb_epoch=10,
verbose=1,
validation_data=(val_x, val_y))
def build_model(self):
inputs = layers.Input(shape=(3, self.sequence_size, self.img_size,
self.img_size))
x = layers.Convolution3D(16,
3,
3,
3,
activation=Relu,
border_mode="same")(inputs)
x = layers.Convolution3D(1,
3,
3,
3,
activation=Relu,
border_mode="same")(x)
print(x.get_shape())
x = layers.Reshape([self.sequence_size, 321, 321])(x)
print(x.get_shape())
x = layers.Convolution2D(64, 3, 3, activation=Relu,
border_mode="same")(x)
x = layers.MaxPooling2D((2, 2))(x)
print(x.get_shape())
x = layers.Convolution2D(32, 3, 3, activation=Relu,
border_mode="same")(x)
x = layers.Convolution2D(32, 3, 3, activation=Relu)(x)
x = layers.Convolution2D(4, 3, 3, activation=Relu,
border_mode="same")(x)
x = layers.Convolution2D(1,
3,
3,
activation='sigmoid',
border_mode="same")(x)
print(x.get_shape())
model = Model(input=inputs, output=x)
model.compile(optimizer='rmsprop',
loss=self.loss_DSSIS_tf11,
metrics=['accuracy'])
return model