def build_model(num_classes, model="B7"):
inputs = layers.Input(shape=(WIDTH, HEIGHT, 3))
if model == "B7":
model = EfficientNetB7(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model == "B6":
model = EfficientNetB6(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model == "B5":
model = EfficientNetB5(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model == "B4":
model = EfficientNetB4(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model == "B3":
model = EfficientNetB3(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model == "B2":
model = EfficientNetB2(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model == "B1":
model = EfficientNetB1(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model == "B0":
model = EfficientNetB0(include_top=False,
input_tensor=inputs,
weights="imagenet")
# Freeze the pretrained weights
model.trainable = False
# Rebuild top
x = layers.GlobalAveragePooling2D(name="avg_pool")(model.output)
x = layers.BatchNormalization()(x)
top_dropout_rate = 0.2
x = layers.Dropout(top_dropout_rate, name="top_dropout", seed=SEED)(x)
outputs = layers.Dense(num_classes,
activation="softmax",
name="predictions")(x)
# Compile
model = tf.keras.Model(inputs, outputs, name="EfficientNet")
optimizer = tf.keras.optimizers.Adam(learning_rate=1e-2)
model.compile(
optimizer=optimizer,
loss="sparse_categorical_crossentropy",
metrics=[tf.keras.metrics.SparseTopKCategoricalAccuracy(k=1)])
return model
def get_parent(input_shape):
if input_shape[0] == 256:
parent = EfficientNetB0(include_top=False)
elif input_shape[0] == 384:
parent = EfficientNetB4(include_top=False)
elif input_shape[0] == 512:
parent = EfficientNetB5(include_top=False)
else:
print('Not a valid input shape for EfficientNet parent')
parent = None
return parent
def model_dense(image_shape=(512, 512, 3)):
in_src = Input(shape=image_shape)
#d = BatchNormalization()(in_src)
m = EfficientNetB5(include_top=False,
weights='imagenet',
input_shape=(512, 512, 3))(in_src)
#Here I am making all the layer of the last layer to be non trainable
# for layer in m.layers[:len(m.layers)-46]:
# layer.trainable = False
#x = tf.keras.layers.GlobalMaxPool2D()(model)
x = tf.keras.layers.GlobalAveragePooling2D()(m)
#x = Flatten()(model)
x = Dense(2048, activation='relu')(x)
x = Dense(1024, activation='relu')(x)
x = Dense(3, activation='softmax')(x)
model = Model(in_src, x)
model.load_weights("saved-model-05-0.35.hdf5")
opt = Adam(lr=0.0002, beta_1=0.5)
model.compile(loss='categorical_crossentropy',
optimizer=opt,
metrics=['accuracy'])
return model
filepath = "best.model.hdf5"
checkpoint = ModelCheckpoint(filepath,
save_best_only = True,
save_weights_only = True,
monitor = 'val_loss',
mode = 'min', verbose = 1)
early_stopping = EarlyStopping(monitor='val_loss', mode="min", verbose=1, patience=15)
reduce_lr = ReduceLROnPlateau(monitor = 'val_loss', factor = 0.3,
patience = 3,
mode = 'min', verbose = 1)
model = Sequential()
model.add(EfficientNetB5(include_top = False, weights = "imagenet",
input_shape=(img_size, img_size, 3)))
model.add(tf.keras.layers.AveragePooling2D(pool_size=(3, 3)))
model.add(tf.keras.layers.Flatten())
model.add(Dropout(0.5))
model.add(tf.keras.layers.Dense(128, activation = "relu"))
model.add(tf.keras.layers.Dense(64, activation = "relu"))
model.add(tf.keras.layers.Dense(5, activation = "softmax"))
model.compile(optimizer = 'adam',
loss = "categorical_crossentropy",
metrics = ["accuracy"])
model.summary()
#class_weights = class_weight.compute_class_weight('balanced', np.unique(train_data.classes),train_data.classes)
class_weights = {0:3.85975197,1: 1.95299487,2: 1.81047065,3: 0.32543726,4: 1.6563135}
shuffle=True)
#control
image_size = (160, 160, 3)
bts = 32
optimizer = Adam(learning_rate=0.001)
train_generator = idg.flow(x_train, y_train, batch_size=bts, seed=2048)
valid_generator = idg2.flow(x_val, y_val)
test_generator = idg2.flow(target)
#2. MODEL
from tensorflow.keras.applications import EfficientNetB5
from tensorflow.keras import regularizers
TF = EfficientNetB5(weights="imagenet",
include_top=False,
input_shape=image_size)
TF.trainable = True
x = TF.output
x = Conv2D(256,
2,
padding='SAME',
activation='swish',
activity_regularizer=regularizers.l2(1e-4),
kernel_regularizer=regularizers.l2(1e-4))(x)
x = GlobalAveragePooling2D()(x)
x = Flatten()(x)
x = Dense(3096, activation='swish')(x)
x = GaussianDropout(rate=0.2)(x)
x_val = scaler.transform(x_val)
#to_categorical
y_train = to_categorical(y_train)
y_test = to_categorical(y_test)
y_val = to_categorical(y_val)
#reshape
x_train = x_train.reshape(-1, 32, 32, 3)
x_test = x_test.reshape(-1, 32, 32, 3)
x_val = x_val.reshape(-1, 32, 32, 3)
print(x_train.shape, x_test.shape, x_val.shape)
#2. 모델링
TF = EfficientNetB5(weights='imagenet',
include_top=False,
input_shape=(32, 32, 3)) #레이어 16개
TF.trainable = False #훈련시키지 않고 가중치만 가져오겠다.
model = Sequential()
model.add(TF)
model.add(Flatten())
model.add(Dense(32, activation='relu'))
model.add(Dense(16, activation='relu'))
model.add(Dense(10,
activation='softmax')) #activation='softmax')) #mnist사용할 경우
model.summary()
print(len(TF.weights)) # 26
print(len(TF.trainable_weights)) # 0
#컴파일, 훈련
from tensorflow.keras.callbacks import EarlyStopping
def create_efficientnet(width, height, depth, model_base,
first_layers_to_freeze, num_classes, learning_rate,
epochs):
inputShape = (height, width, depth)
inputs = K.Input(shape=inputShape)
if model_base == "b0":
effnet = EfficientNetB0(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b1":
effnet = EfficientNetB1(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b2":
effnet = EfficientNetB2(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b3":
effnet = EfficientNetB3(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b4":
effnet = EfficientNetB4(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b5":
effnet = EfficientNetB5(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b6":
effnet = EfficientNetB6(include_top=False,
input_tensor=inputs,
weights="imagenet")
else:
effnet = EfficientNetB7(include_top=False,
input_tensor=inputs,
weights="imagenet")
# # Print architecture of effnet
# for i, layer in enumerate(effnet.layers[:]):
# print(i, layer.name, layer.output_shape)
# print(f"Effnet len: {len(effnet.layers[:])}")
# b0: 20; b2: 33; b4: 236; b6: 45; b7: 265
for i, layer in enumerate(effnet.layers[:first_layers_to_freeze]):
layer.trainable = False
for i, layer in enumerate(effnet.layers[first_layers_to_freeze:]):
if not isinstance(layer, K.layers.BatchNormalization):
layer.trainable = True
model = Sequential()
model.add(effnet)
model.add(K.layers.Dropout(0.25))
model.add(K.layers.Dense(effnet.layers[-1].output_shape[3]))
model.add(K.layers.LeakyReLU())
model.add(K.layers.GlobalAveragePooling2D())
model.add(K.layers.BatchNormalization())
model.add(K.layers.Dropout(0.5))
model.add(K.layers.Dense(num_classes, activation='softmax'))
# Freeze the batchnorm layer of our model
for i, layer in enumerate(model.layers[:]):
if isinstance(layer, K.layers.BatchNormalization):
layer.trainable = False
opt = Adam(lr=learning_rate, decay=learning_rate / epochs)
model.compile(loss="categorical_crossentropy",
optimizer=opt,
metrics=["accuracy"])
model.summary()
return model
train_size=0.9,
random_state=128,
shuffle=True)
#control
bts = 128
optimizer = Adam(learning_rate=1e-3)
train_generator = idg.flow(x_train, y_train, batch_size=bts, seed=2048)
valid_generator = idg2.flow(x_val, y_val)
test_generator = idg2.flow(target)
#2. MODEL
from tensorflow.keras.applications import EfficientNetB5
TF = EfficientNetB5(include_top=False,
weights='imagenet',
input_shape=x_train.shape[1:])
TF.trainable = True
x = TF.output
x = GlobalAveragePooling2D()(x)
x = Flatten()(x)
x = Dense(2048, activation='relu')(x)
x = Dropout(0.3)(x)
x = Dense(1000, activation='softmax')(x)
model = Model(inputs=TF.input, outputs=x)
model.summary()
#COMPILE
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau
model.compile(loss='categorical_crossentropy',
optimizer=optimizer,
training_datagen = ImageDataGenerator(rotation_range=10,brightness_range=(1.1, 0.9),shear_range=10.0,zoom_range=[0.8, 1.2],vertical_flip=True,fill_mode='wrap',
width_shift_range=0.2,
height_shift_range=0.2,
horizontal_flip=True)
valid_datagen = ImageDataGenerator()
t_gen = training_datagen.flow(X_train, y_train, batch_size=4)
v_gen = valid_datagen.flow(X_test, y_test, batch_size=4)
#t_gen = training_datagen.flow(X_train, y_train)
#v_gen = valid_datagen.flow(X_test, y_test)
image_input = tf.keras.Input(shape=(456, 456, 3))
#model = ResNet50(weights='imagenet')
model = EfficientNetB5(input_tensor=image_input, include_top=False,weights="imagenet")
model.summary()
model.trainable = False
last_layer = model.output
last_layer = tf.keras.layers.GlobalAveragePooling2D(name= 'avg_pool_last')(last_layer)
#x= tf.keras.layers.Flatten(name='flatten')(last_layer)
last_layer = tf.keras.layers.BatchNormalization(name = 'batch_last')(last_layer)
top_dropout_rate = 0.2
last_layer = tf.keras.layers.Dropout(top_dropout_rate, name = 'top_dropout')(last_layer)
out = tf.keras.layers.Dense(num_classes, activation='softmax', name='output_layer')(last_layer)
#image_input = Input(shape=(500, 500, 3))
custom_resnet_model = tf.keras.Model(inputs=image_input,outputs= out)
custom_resnet_model.summary()
#for layer in custom_resnet_model.layers[:-4]:
# layer.trainable = False
def create_efficientnet(width, height, depth, model_base,
first_layers_to_freeze):
inputShape = (height, width, depth)
inputs = K.Input(shape=inputShape)
if model_base == "b0":
effnet = EfficientNetB0(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b1":
effnet = EfficientNetB1(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b2":
effnet = EfficientNetB2(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b3":
effnet = EfficientNetB3(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b4":
effnet = EfficientNetB4(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b5":
effnet = EfficientNetB5(include_top=False,
input_tensor=inputs,
weights="imagenet")
elif model_base == "b6":
effnet = EfficientNetB6(include_top=False,
input_tensor=inputs,
weights="imagenet")
else:
effnet = EfficientNetB7(include_top=False,
input_tensor=inputs,
weights="imagenet")
# # Print architecture of effnet
# for i, layer in enumerate(effnet.layers[:]):
# print(i, layer.name, layer.output_shape)
# b0: 20; b2: 33; b4: 147; b6: 45; b7: 265
for i, layer in enumerate(effnet.layers[:first_layers_to_freeze]):
layer.trainable = False
for i, layer in enumerate(effnet.layers[first_layers_to_freeze:]):
layer.trainable = True
effnet.summary()
model = Sequential()
model.add(effnet)
model.add(K.layers.Dropout(0.25))
model.add(K.layers.Dense(effnet.layers[-1].output_shape[3]))
model.add(K.layers.LeakyReLU())
model.add(K.layers.GlobalAveragePooling2D())
model.add(K.layers.Dropout(0.5))
model.add(K.layers.Dense(1, activation='linear'))
return model