def build_cnn(l1,l2):
model = Sequential()
# logpolar layer
if l1 == 'LPC':
model.add(LogPolar(input_shape=(h, w, dp))) # 200x150x3 => 200x150x3
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2))) # 200x150x3 => 100x75
model.add(Dropout(0.25))
if l1 == 'LP':
model.add(LogPolar(input_shape=(h, w, dp))) # 200x150x3 => 200x150x3
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2))) # 200x150x3 => 100x75
model.add(Dropout(0.25))
if l1 == 'C':
model.add(Conv2D(32, (3, 3), activation='relu', input_shape=(h, w, dp))) # 200x150x3 => 198x148x32
model.add(Conv2D(32, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2))) # 198x148x32 => 99x74x32
model.add(Dropout(0.25))
if l2 == 'LPC':
model.add(LogPolar()) # 200x150x3 => 200x150x3
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2))) # 200x150x3 => 100x75
model.add(Dropout(0.25))
if l2 == 'LP':
model.add(LogPolar()) # 99x74x32 => 99x74x32
model.add(Conv2D(64, (3, 3), activation='relu'))
#model.add(Lambda(LogPolarLambda, output_shape=LogPolarLambda_output_shape)) # 99x74x32 => 99x74x32
model.add(MaxPooling2D(pool_size=(2, 3))) # 99x74x32 => 33x37x32
model.add(Dropout(0.25))
if l2 == 'C':
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(Conv2D(64, (3, 3), activation='relu'))
model.add(MaxPooling2D(pool_size=(2, 2))) # 200x150 => 100x75
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(1024, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(n_classes, activation='softmax'))
# Compile model
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
return model
def build_cnn_alex(l1, l2):
""""
The first convolutional layer filters the 224×224×3 input image with 96 kernels of size 11×11×3
with a stride of 4 pixels (this is the distance between the receptive field centers of neighboring
neurons in a kernel map). The second convolutional layer takes as input the (response-normalized
and pooled) output of the first convolutional layer and filters it with 256 kernels of size 5 × 5 × 48.
The third, fourth, and fifth convolutional layers are connected to one another without any intervening
pooling or normalization layers. The third convolutional layer has 384 kernels of size 3 × 3 ×
256 connected to the (normalized, pooled) outputs of the second convolutional layer. The fourth
convolutional layer has 384 kernels of size 3 × 3 × 192 , and the fifth convolutional layer has 256
kernels of size 3 × 3 × 192. The fully-connected layers have 4096 neurons each.
https://papers.nips.cc/paper/4824-imagenet-classification-with-deep-convolutional-neural-networks.pdf
"""
# Define model architecture
model = Sequential()
#AlexNet with batch normalization in Keras
#input image is 224x224
if l1 == 'LPC':
model.add(LogPolar(input_shape=(h, w, dp))) # 200x150x3 => 200x150x3
model.add(Conv2D(64, (11, 11)))
if l1 == 'LP':
model.add(LogPolar(input_shape=(h, w, dp))) # 200x150x3 => 200x150x3
if l1 == 'C':
model.add(Conv2D(64, (11, 11), input_shape=(h, w, dp)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(3, 3)))
if l2 == 'LPC':
model.add(LogPolar()) # 200x150x3 => 200x150x3
model.add(Conv2D(128, (7, 7)))
if l2 == 'LP':
model.add(LogPolar()) # 200x150x3 => 200x150x3
if l2 == 'C':
model.add(Conv2D(128, (7, 7)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(3, 3)))
model.add(Conv2D(192, (3, 3)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(3, 3)))
model.add(Conv2D(256, (3, 3)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(3, 3)))
model.add(Flatten())
model.add(Dense(4096, kernel_initializer='normal'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dense(4096, kernel_initializer='normal'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dense(n_classes, kernel_initializer='normal'))
model.add(BatchNormalization())
model.add(Activation('softmax'))
# Compile model
sgd = optimizers.SGD(lr=0.01, momentum=0.9, decay=0.0005)
model.compile(loss='categorical_crossentropy',
optimizer=sgd,
metrics=['accuracy'])
return model