def build_model_kernel_cascade753():
input_shape_img = utils.get_rgb_input_shape(IMG_ROWS, IMG_COLUMNS)
num_classes = 2
large_kernel_shape = (7, 7)
medium_kernel_shape = (5, 5)
small_kernel_shape = (3, 3)
model = Sequential()
model.add(
Conv2D(48,
kernel_size=large_kernel_shape,
activation='relu',
padding='same',
input_shape=input_shape_img))
model.add(
Conv2D(48,
kernel_size=large_kernel_shape,
activation='relu',
padding='same'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(
Conv2D(64,
kernel_size=medium_kernel_shape,
activation='relu',
padding='same'))
model.add(
Conv2D(64,
kernel_size=medium_kernel_shape,
activation='relu',
padding='same'))
model.add(Conv2D(32, kernel_size=(1, 1), activation='relu',
padding='same'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(
Conv2D(16,
kernel_size=small_kernel_shape,
activation='relu',
padding='same'))
model.add(
Conv2D(16,
kernel_size=small_kernel_shape,
activation='relu',
padding='same'))
model.add(Conv2D(8, kernel_size=(1, 1), activation='relu', padding='same'))
model.add(GlobalAveragePooling2D())
model.add(Dense(num_classes, activation='softmax'))
model.compile(loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.Adam(),
metrics=['accuracy'])
model_name = "fcn_kernel_cascade753"
return (model, model_name)
def build_model_3modules_cascade753():
input_shape_img = utils.get_rgb_input_shape(IMG_ROWS, IMG_COLUMNS)
block1_units = 48
block2_units = 60
block3_units = 72
block1_kernel_shape = (7, 7)
block2_kernel_shape = (5, 5)
block3_kernel_shape = (3, 3)
model = Sequential()
model.add(Conv2D(block1_units, kernel_size=block1_kernel_shape,
activation='relu', input_shape=input_shape_img,
name='input_layer'))
model.add(Conv2D(block1_units, kernel_size=block1_kernel_shape,
activation='relu', padding='same'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(block2_units, kernel_size=block2_kernel_shape,
activation='relu', padding='same'))
model.add(Conv2D(block2_units, kernel_size=block2_kernel_shape,
activation='relu', padding='same'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Conv2D(block3_units, kernel_size=block3_kernel_shape,
activation='relu', padding='same'))
model.add(Conv2D(block3_units, kernel_size=block3_kernel_shape,
activation='relu', padding='same'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))
model.add(Flatten())
model.add(Dense(DENSE_UNITS, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(NUM_CLASSES, activation='softmax',
name='logits_layer'))
model.compile(loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.Adam(),
metrics=['accuracy'])
model_name = "vgg_3modules_cascade753_dense{}".format(DENSE_UNITS)
return (model, model_name)