MNIST 실습 - CNN 모델

import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
 
mnist = input_data.read_data_sets("./mnist/data/", one_hot=True)
 
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense
 
 
 
# 레이어 설정
X = tf.placeholder(tf.float32, shape=[None, 28, 28, 1])
Y = tf.placeholder(tf.float32, [None, 10])
 
W1 = tf.Variable(tf.random_normal([4,4,1,16], stddev=0.01))
L1 = tf.nn.conv2d(X, W1, strides=[1,1,1,1], padding='SAME')
L1 = tf.nn.relu(L1)
L1 = tf.nn.max_pool(L1, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')
 
W2 = tf.Variable(tf.random_normal([4,4,16,32], stddev=0.01))
L2 = tf.nn.conv2d(L1, W2, strides=[1,1,1,1], padding='SAME')
L2 = tf.nn.relu(L2)
L2 = tf.nn.max_pool(L2, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')
 
L3 = tf.reshape(L2, [-1, 7*7*32])
W3 = tf.Variable(tf.random_normal([7*7*32, 256], stddev=0.01))
L3 = tf.matmul(L3, W3)
L = tf.nn.relu(L3)
#L3 = tf.nn.dropout(L3,keep_prob=0.5)
 
W4 = tf.Variable(tf.random_normal([256, 10], stddev=0.01))
model = tf.matmul(L3, W4)
model = tf.nn.softmax(model)
 
 
 
# 손실함수, 옵티마이저
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=model, labels=Y),axis=0)
optimizer = tf.train.AdamOptimizer(0.01).minimize(cost)
 
 
 
# 정확도 
correct = tf.equal(tf.arg_max(model, 1), tf.arg_max(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32),axis=0)
 
 
 
import random
 
epoch = 5
batch_s = 1000
batch_num = int(mnist.train.num_examples/batch_s)
 
with tf.Session() as sess:
  initial = tf.global_variables_initializer()
  sess.run(initial)
 
  # 학습 과정
  for i in range(epoch):    
    avg_cost=0
    
    for j in range(batch_num):
      batch_X, batch_Y = mnist.train.next_batch(batch_s)
      batch_X = batch_X.reshape(-1, 28, 28, 1)
      
      tmp_cost, opt = sess.run([cost, optimizer], feed_dict={X: batch_X, Y: batch_Y})
      avg_cost += tmp_cost
      
    avg_cost = avg_cost/batch_num
    print("epoch: ", i+1, "cost >> ", avg_cost)
 
 
  # 정확도 출력하기  
  input_test = mnist.test.images
  output_test = mnist.test.labels
 
  input_test = input_test.reshape(-1, 28, 28, 1)
  
  acc = sess.run(accuracy, feed_dict={X: input_test, Y: output_test})
  print("정확도 : ", acc)
 
 
  # 실제 라벨 예측하기
  idx = random.randint(0, mnist.test.num_examples-1)
 
  predict_X = mnist.test.images[idx, idx+1]
  predict_X = predict_X.reshape(-1, 28, 28, 1)
  predict_X = sess.run(model, feed_dict={X: predict_X})
 
  predict_Y = mnist.test.labels[idx, idx+1]
 
  label_real = tf.arg_max(predict_Y, 1)
  label_predict = tf.arg_max(predict_X, 1)
 
  print("실제 라벨 값 : ", label_real)
  print("예측 라벨 값 : ", label_predict)