°øÀ¯ÀÚ·á HOME > ÀÚ·á½Ç > °øÀ¯ÀÚ·á
 
[PYTHON] MNIST À̹ÌÁö ÆÇÁ¤ÇÏ´Â CNN ÆÄÀ̼±
°ü¸®ÀÚ 19-10-23 11:19 3,078

from keras.datasets import mnist
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()


print(train_images.shape)  # (60000, 28, 28)
print(test_images.shape)   # (10000, 28, 28)


print(len(train_labels))  # 60000
print(len(test_labels))  # 10000


train_labels    # array([5, 0, 4, ..., 5, 6, 8], dtype=uint8)
test_labels    # array([7, 2, 1, ..., 4, 5, 6], dtype=uint8)


from keras import models
from keras import layers
network = models.Sequential()                               # ¸ðµ¨ ÃʱâÈ­


# Àº´ÐÃþ ¼³Á¤. Data shape°¡ 28*28À̹ǷΠÀÌ¿¡ ¸ÂÃß¾î input_shape¸¦ ³Ö´Â´Ù
network.add(layers.Dense(512, activation='relu', input_shape=(28*28,)))


# Ãâ·ÂÃþ ¼³Á¤. 10°³ °¢°¢¿¡ ´ëÇÑ È®·üÁ¤º¸ Ãâ·Â. Softmax ÃþÀº È®·ü Á¡¼ö¸¦ Ãâ·ÂÇÑ´Ù
network.add(layers.Dense(10, activation='softmax'))

network.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])
print(train_images[0,])


## reshapeÇϱâ
train_images = train_images.reshape((60000, 28*28))
train_images = train_images.astype('float32')/255


test_images = test_images.reshape((10000, 28*28))
test_images = test_images.astype('float32')/255


## ÆǺ°À» ÇÏ·Á¸é Á¾¼Óº¯¼ö°¡ ¸í¸ñº¯¼ö°¡ µÇ¾î¾ß ÇÔ
from tensorflow.keras.utils import to_categorical
train_labels = to_categorical(train_labels)
test_labels = to_categorical(test_labels)


## ÇнÀÇϱâ
network.fit(train_images, train_labels, epochs=5, batch_size=128)


## ÇнÀµÈ ¸ðµ¨À» °¡Áö°í ¿¹ÃøÇϱâ
test_loss, test_acc = network.evaluate(test_images, test_labels)
print('test_acc:', test_acc)