迁移学习，微调和tensorflow-hub

Pretrained-Model

(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train = x_train.reshape(-1, 28, 28, 1).astype('float32') / 255.0
x_test = x_test.reshape(-1, 28, 28, 1).astype('float32') / 255.0

model = keras.models.load_model('pretrained/')
print(model.summary())
model.trainable = False

for layer in model.layers:
    assert layer.trainable == False
    layer.trainable = False

base_inputs = model.layers[0].input
base_outputs = model.layers[-2].output
final_outputs = layers.Dense(10)(base_outputs)

new_model = keras.Model(inputs=base_inputs, outputs=final_outputs)
# print(new_model.summary())
new_model.compile(
    optimizer=keras.optimizers.Adam(),
    loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),
    metrics=['accuracy'],
)

Pretrained Keras Model

x = tf.random.normal(shape=(5, 299, 299, 3))
y = tf.constant([0, 1, 2, 3, 4])

model = keras.applications.InceptionV3(include_top=True)
# print(model.summary())
base_inputs = model.layers[0].input
base_outputs = model.layers[-2].output
final_outputs = layers.Dense(5)(base_outputs)
new_model = keras.Model(inputs=base_inputs, outputs=final_outputs)

new_model.compile(
    optimizer=keras.optimizers.Adam(),
    loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),
    metrics=['accuracy'],
)

new_model.fit(x, y, epochs=15, verbose=2)

Pretrained Hub Model

x = tf.random.normal(shape=(5, 299, 299, 3))
y = tf.constant([0, 1, 2, 3, 4])

url = 'https://tfhub.dev/google/imagenet/inception_v3/feature_vector/4'

base_model = hub.KerasLayer(url, input_shape=(299, 299, 3))
base_model.trainable = False
model = keras.Sequential([
    base_model,
    layer.Dense(128, activation='relu'),
    layer.Dense(64, activation='relu'),
    layer.Dense(5),
])

model.compile(
    optimizer=keras.optimizers.Adam(),
    loss=keras.losses.SparseCategoricalCrossentropy(from_logits=True),
    metrics=['accuracy'],
)

model.fit(x, y, batch_size=32, epochs=15, verbose=2)