def mldensenet(img_shape, n_classes, mltype=0, finalAct='softmax', f=32):
"""
if mltype == 0:
repetitions = 6,12,24,16
elif mltype == 2 or mltype == 3:
repetitions = 6,12,24,16
elif mltype == 1 or mltype == 4 or mltype == 5:
repetitions = 6,12,24
"""
if mltype == 0:
repetitions = 6, 12, 48, 32
elif mltype == 2 or mltype == 3:
repetitions = 6, 12, 48, 32
elif mltype == 1 or mltype == 4 or mltype == 5:
repetitions = 6, 12, 32
def bn_rl_conv(x, f, k=1, s=1, p='same'):
x = BatchNormalization(epsilon=1.001e-5)(x)
x = ReLU()(x)
x = Conv2D(f, k, strides=s, padding=p)(x)
return x
def dense_block(tensor, r):
for _ in range(r):
x = bn_rl_conv(tensor, 4 * f)
x = bn_rl_conv(x, f, 3)
x = squeeze_excite_block(x)
tensor = Concatenate()([tensor, x])
return tensor
def transition_block(x):
x = bn_rl_conv(x, K.int_shape(x)[-1] // 2)
#x = Dropout(0.2)(x)
x = AvgPool2D(2, strides=2, padding='same')(x)
return x
def squeeze_excite_block(tensor, ratio=16):
init = tensor
channel_axis = 1 if K.image_data_format() == "channels_first" else -1
filters = init.shape[channel_axis]
se_shape = (1, 1, filters)
se = GlobalAvgPool2D()(init)
se = Reshape(se_shape)(se)
se = Dense(filters // ratio,
activation='relu',
kernel_initializer='he_normal',
use_bias=False)(se)
se = Dense(filters,
activation='sigmoid',
kernel_initializer='he_normal',
use_bias=False)(se)
if K.image_data_format() == 'channels_first':
se = Permute((3, 1, 2))(se)
x = multiply([init, se])
return x
input = Input(img_shape)
x = Conv2D(64, 7, strides=2, padding='same')(input)
x = BatchNormalization(epsilon=1.001e-5)(x)
x = ReLU()(x)
x = MaxPool2D(3, strides=2, padding='same')(x)
for r in repetitions:
d = dense_block(x, r)
#d = squeeze_excite_block(d)
x = transition_block(d)
if mltype == 0:
x = GlobalAvgPool2D()(d)
output = Dense(n_classes, activation=finalAct)(x)
model = Model(input, output)
return model
elif mltype == 1:
outputs = []
mt_block = 16
for i in range(n_classes):
print("class ", i)
d = dense_block(x, mt_block)
branch = GlobalAvgPool2D()(d)
output = Dense(1, activation=finalAct)(branch)
outputs.append(output)
outputs = Concatenate()(outputs)
model = Model(input, outputs)
return model
elif mltype == 2:
x = GlobalAvgPool2D()(d)
outputs = []
for i in range(n_classes):
print("\rclass ", i, end="")
x = Dense(1024, activation='relu')(x)
x = Dense(512, activation='relu')(x)
x = Dense(256, activation='relu')(x)
output = Dense(1, activation=finalAct)(x)
outputs.append(output)
outputs = Concatenate()(outputs)
model = Model(input, outputs)
return model
elif mltype == 3:
a = transition_block(d)
outputs = []
for i in range(n_classes):
print("\rclass ", i, end="")
v = Conv2D(256, (3, 3), activation='relu', padding='same')(a)
v = Conv2D(256, (3, 3), activation='relu', padding='same')(v)
v = Conv2D(256, (3, 3), activation='relu', padding='same')(v)
v = MaxPooling2D((2, 2), strides=(2, 2))(v)
v = Conv2D(512, (3, 3), activation='relu', padding='same')(v)
v = Conv2D(512, (3, 3), activation='relu', padding='same')(v)
v = Conv2D(512, (3, 3), activation='relu', padding='same')(v)
v = MaxPooling2D((2, 2), strides=(2, 2))(v)
v = Flatten()(v)
d = Dense(4096, activation='relu')(v)
d = Dense(4096, activation='relu')(d)
d = Dense(1024, activation='relu')(d)
output = Dense(1, activation=finalAct)(d)
outputs.append(output)
outputs = Concatenate()(outputs)
model = Model(input, outputs)
return model
elif mltype == 4:
x = GlobalAvgPool2D()(d)
outputs = []
for i in range(n_classes):
print("\rclass ", i, end="")
x = Dense(1024, activation='relu')(x)
x = Dense(512, activation='relu')(x)
x = Dense(256, activation='relu')(x)
output = Dense(1, activation=finalAct)(x)
outputs.append(output)
outputs = Concatenate()(outputs)
model = Model(input, outputs)
return model
elif mltype == 5:
a = transition_block(d)
outputs = []
for i in range(n_classes):
print("\rclass ", i, end="")
v = Conv2D(256, (3, 3), activation='relu', padding='same')(a)
v = Conv2D(256, (3, 3), activation='relu', padding='same')(v)
v = Conv2D(256, (3, 3), activation='relu', padding='same')(v)
v = MaxPooling2D((2, 2), strides=(2, 2))(v)
v = Conv2D(512, (3, 3), activation='relu', padding='same')(v)
v = Conv2D(512, (3, 3), activation='relu', padding='same')(v)
v = Conv2D(512, (3, 3), activation='relu', padding='same')(v)
v = MaxPooling2D((2, 2), strides=(2, 2))(v)
v = Flatten()(v)
d = Dense(2048, activation='relu')(v)
d = Dense(2048, activation='relu')(d)
d = Dense(1024, activation='relu')(d)
output = Dense(1, activation=finalAct)(d)
outputs.append(output)
outputs = Concatenate()(outputs)
model = Model(input, outputs)
return model
