import tensorflow as tf
from tensorflow.keras import layers
from tensorflow.keras import datasets
input_shape = (28, 28, 1)
num_classes = 10
inputs = layers.Input(shape = input_shape)
# Feature Extraction
# 1st Block
net = layers.Conv2D(32, 3, padding='SAME')(inputs)
net = layers.Activation('relu')(net)
net = layers.Conv2D(32, 3, padding='SAME')(net)
net = layers.Activation('relu')(net)
net = layers.MaxPool2D((2,2))(net)
net = layers.Dropout(0.25)(net)
#2nd Block
net = layers.Conv2D(64, 3, padding='SAME')(net)
net = layers.Activation('relu')(net)
net = layers.Conv2D(64, 3, padding='SAME')(net)
net = layers.Activation('relu')(net)
net = layers.MaxPool2D((2,2))(net)
net = layers.Dropout(0.25)(net)
# Fully Connected
net = layers.Flatten()(net)
net = layers.Dense(512)(net)
net = layers.Activation('relu')(net)
net = layers.Dropout(0.25)(net)
net = layers.Dense(num_classes)(net)
#이때 num_classes는 데이터의 클래스 개수로 해야 된다.(노드의 개수)
net = layers.Activation('softmax')(net)
#softmax는 확률로 나타내는 것을 말한다.
model = tf.keras.Model(inputs = inputs, outputs = net, name = 'CNN_Basic')
Expert한 방법
mnist = tf.keras.datasets.mnist
# Load data
(x_train, y_train), (x_test, y_test) = mnist.load_data()
# 차원 추가
x_train = x_train[..., tf.newaxis]
x_test = x_train[..., tf.newaxis]
# Data Normalization
x_train, x_test = x_train/255., x_test/255.
train_ds = tf.data.Datasets.from_tensor_slices((x_train, y_train))
train_ds = train_ds.shuffle(1000) #1000은 버퍼 사이즈. 1000이 가장 적절함. 디폴드 값이라 생각
train_ds = train_ds.batch(32)
test_ds = tf.data.Dataset.from_tensor_slices((x_test, y_test))
test_ds = test_ds.batch(32)