from PrognosAIs.Model.Architectures.Architecture import ClassificationNetworkArchitecture
from tensorflow.keras.layers import Conv2D
from tensorflow.keras.layers import Conv3D
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import Dropout
from tensorflow.keras.layers import Flatten
from tensorflow.keras.layers import MaxPooling2D
from tensorflow.keras.layers import MaxPooling3D
from tensorflow.keras.layers import ReLU
from tensorflow.keras.models import Model
[docs]class AlexNet_2D(ClassificationNetworkArchitecture):
padding_type = "valid"
[docs] def create_model(self):
self.check_minimum_input_size(self.inputs, [67, 67])
stride_size = self.get_corrected_stride_size(self.inputs, [4, 4], [11, 11])
conv_1 = Conv2D(filters=96, kernel_size=(11, 11), strides=stride_size)(self.inputs)
relu_1 = ReLU()(conv_1)
stride_size = self.get_corrected_stride_size(relu_1, [2, 2], [3, 3])
pooling_1 = MaxPooling2D(
pool_size=(3, 3),
strides=stride_size,
padding=self.padding_type,
data_format="channels_last",
)(relu_1)
conv_2 = Conv2D(filters=256, kernel_size=(5, 5), strides=(1, 1), padding="same")(pooling_1)
relu_2 = ReLU()(conv_2)
stride_size = self.get_corrected_stride_size(relu_2, [2, 2], [3, 3])
pooling_2 = MaxPooling2D(
pool_size=(3, 3),
strides=stride_size,
padding=self.padding_type,
data_format="channels_last",
)(relu_2)
conv_3 = Conv2D(filters=384, kernel_size=(3, 3), strides=(1, 1), padding="same")(pooling_2)
relu_3 = ReLU()(conv_3)
conv_4 = Conv2D(filters=384, kernel_size=(3, 3), strides=(1, 1), padding="same")(relu_3)
relu_4 = ReLU()(conv_4)
conv_5 = Conv2D(filters=256, kernel_size=(3, 3), strides=(1, 1), padding="same")(relu_4)
relu_5 = ReLU()(conv_5)
stride_size = self.get_corrected_stride_size(relu_5, [2, 2], [3, 3])
pooling_5 = MaxPooling2D(
pool_size=(3, 3),
strides=stride_size,
padding=self.padding_type,
data_format="channels_last",
)(relu_5)
flatten_1 = Flatten()(pooling_5)
dense_1 = Dense(
units=4096,
activation=None,
use_bias=True,
kernel_initializer="glorot_uniform",
bias_initializer="zeros",
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
)(flatten_1)
dropout_1 = Dropout(0.5)(dense_1)
dense_2 = Dense(
units=4096,
activation=None,
use_bias=True,
kernel_initializer="glorot_uniform",
bias_initializer="zeros",
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
)(dropout_1)
dropout_2 = Dropout(0.5)(dense_2)
predictions = self.outputs(dropout_2)
model = Model(inputs=self.inputs, outputs=predictions)
return model
[docs]class AlexNet_3D(ClassificationNetworkArchitecture):
padding_type = "valid"
[docs] def create_model(self):
self.check_minimum_input_size(self.inputs, [67, 67, 67])
stride_size = self.get_corrected_stride_size(self.inputs, [4, 4, 4], [11, 11, 11])
conv_1 = Conv3D(filters=96, kernel_size=(11, 11, 11), strides=stride_size)(self.inputs)
relu_1 = ReLU()(conv_1)
stride_size = self.get_corrected_stride_size(relu_1, [2, 2, 2], [3, 3, 3])
pooling_1 = MaxPooling3D(
pool_size=(3, 3, 3),
strides=stride_size,
padding=self.padding_type,
data_format="channels_last",
)(relu_1)
conv_2 = Conv3D(filters=256, kernel_size=(5, 5, 5), strides=(1, 1, 1), padding="same")(
pooling_1
)
relu_2 = ReLU()(conv_2)
stride_size = self.get_corrected_stride_size(relu_2, [2, 2, 2], [3, 3, 3])
pooling_2 = MaxPooling3D(
pool_size=(3, 3, 3),
strides=stride_size,
padding=self.padding_type,
data_format="channels_last",
)(relu_2)
conv_3 = Conv3D(filters=384, kernel_size=(3, 3, 3), strides=(1, 1, 1), padding="same")(
pooling_2
)
relu_3 = ReLU()(conv_3)
conv_4 = Conv3D(filters=384, kernel_size=(3, 3, 3), strides=(1, 1, 1), padding="same")(
relu_3
)
relu_4 = ReLU()(conv_4)
conv_5 = Conv3D(filters=256, kernel_size=(3, 3, 3), strides=(1, 1, 1), padding="same")(
relu_4
)
relu_5 = ReLU()(conv_5)
stride_size = self.get_corrected_stride_size(relu_5, [2, 2, 2], [3, 3, 3])
pooling_5 = MaxPooling3D(
pool_size=(3, 3, 3),
strides=stride_size,
padding=self.padding_type,
data_format="channels_last",
)(relu_5)
flatten_1 = Flatten()(pooling_5)
dense_1 = Dense(
units=4096,
activation=None,
use_bias=True,
kernel_initializer="glorot_uniform",
bias_initializer="zeros",
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
)(flatten_1)
dropout_1 = Dropout(0.5)(dense_1)
dense_2 = Dense(
units=4096,
activation=None,
use_bias=True,
kernel_initializer="glorot_uniform",
bias_initializer="zeros",
kernel_regularizer=None,
bias_regularizer=None,
activity_regularizer=None,
kernel_constraint=None,
bias_constraint=None,
)(dropout_1)
dropout_2 = Dropout(0.5)(dense_2)
predictions = self.outputs(dropout_2)
model = Model(inputs=self.inputs, outputs=predictions)
return model