import keras
import keras.backend as K
from keras.layers import Dense, Conv2D, Input, MaxPooling2D, Flatten
from keras.models import Model
from keras.datasets import fashion_mnist
from keras.callbacks import ModelCheckpoint
# setup parameters
batch_sz = 128
nb_class = 10
nb_epochs = 10
img_h, img_w = 28, 28
print( K.image_data_format())
# input image dimensions
img_rows, img_cols = 28, 28
def get_dataset():
"""
Return processed and reshaped dataset for training
In this cases Fashion-mnist dataset.
"""
# load mnist dataset
(x_train, y_train), (x_test, y_test) = fashion_mnist.load_data()
# test and train datasets
print("Nb Train:", x_train.shape[0], "Nb test:",x_test.shape[0])
x_train = x_train.reshape(x_train.shape[0], img_h, img_w, 1)
x_test = x_test.reshape(x_test.shape[0], img_h, img_w, 1)
in_shape = (img_h, img_w, 1)
# normalize inputs
x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255.0
x_test /= 255.0
# convert to one hot vectors
y_train = keras.utils.to_categorical(y_train, nb_class)
y_test = keras.utils.to_categorical(y_test, nb_class)
return x_train, x_test, y_train, y_test
x_train, x_test, y_train, y_test = get_dataset()
def create_model(img_h=28, img_w=28):
inputs = Input(shape=(img_h, img_w, 1))
x = Conv2D(32, kernel_size=(3,3), activation='relu')(inputs) # 32C 3K 1S VP RELU
x = Conv2D(32, kernel_size=(3,3), activation='relu')(x) # 64C 3K 1S VP RELU
x = MaxPooling2D(pool_size=(2,2))(x) # pool2
x = Conv2D(64, kernel_size=(3,3), activation='relu')(x) # 32C 3K 1S VP RELU
x = Conv2D(64, kernel_size=(3,3), activation='relu')(x) # 64C 3K 1S VP RELU
x = MaxPooling2D(pool_size=(2,2))(x) # pool2
x = Flatten()(x)
preds = Dense(nb_class, activation='softmax')(x)
model = Model(inputs=inputs, outputs=preds)
print(model.summary())
return model
model = create_model()
model.compile(loss=keras.losses.categorical_crossentropy,
optimizer=keras.optimizers.SGD(lr=0.001),
metrics=['accuracy'])
callback = ModelCheckpoint()
# start training
model.fit(x_train, y_train,
batch_size=batch_sz,
epochs=nb_epochs,
verbose=1,
validation_data=(x_test, y_test), callbacks=[callback])
# Evaluate
score = model.evaluate(x_test, y_test, verbose=0)
print('Test loss:', score[0])
print('Test accuracy:', score[1])
