[tensorflow2.x 기초 - 6] MNIST data로 Evaluation & Prediction을 구현기본

 

 

 

 

 

TensorFlow: Evaluating & Prediction¶

In [1]:

import tensorflow as tf
from tensorflow.keras import layers

from tensorflow.keras import datasets 

 

Build Model¶

In [2]:

input_shape = (28, 28, 1)
num_classes = 10

learning_rate = 0.001

In [3]:

inputs = layers.Input(input_shape)
net = layers.Conv2D(32, (3, 3), padding='SAME')(inputs)
net = layers.Activation('relu')(net)
net = layers.Conv2D(32, (3, 3), padding='SAME')(net)
net = layers.Activation('relu')(net)
net = layers.MaxPooling2D(pool_size=(2, 2))(net)
net = layers.Dropout(0.5)(net)

net = layers.Conv2D(64, (3, 3), padding='SAME')(net)
net = layers.Activation('relu')(net)
net = layers.Conv2D(64, (3, 3), padding='SAME')(net)
net = layers.Activation('relu')(net)
net = layers.MaxPooling2D(pool_size=(2, 2))(net)
net = layers.Dropout(0.5)(net)

net = layers.Flatten()(net)
net = layers.Dense(512)(net)
net = layers.Activation('relu')(net)
net = layers.Dropout(0.5)(net)
net = layers.Dense(num_classes)(net)
net = layers.Activation('softmax')(net)

model = tf.keras.Model(inputs=inputs, outputs=net, name='Basic_CNN')

In [4]:

# Model is the full model w/o custom layers
model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate),
              loss='sparse_categorical_crossentropy',
              metrics=['accuracy'])

 

Preprocess¶

 

데이터셋 불러오기

In [5]:

(train_x, train_y), (test_x, test_y) = datasets.mnist.load_data()

In [6]:

train_x = train_x[..., tf.newaxis]
test_x = test_x[..., tf.newaxis]

train_x = train_x / 255.
test_x = test_x / 255.

 

Training¶

In [7]:

num_epochs = 1
batch_size = 64

In [8]:

hist = model.fit(train_x, train_y, 
                 batch_size=batch_size, 
                 shuffle=True)

 

938/938 [==============================] - 132s 141ms/step - loss: 0.2164 - accuracy: 0.9312

In [25]:

hist.history

Out[25]:

{'loss': [0.21658981240888436], 'accuracy': [0.931]}

 

Evaluating¶

• 학습한 모델 확인

In [9]:

model.evaluate(test_x, test_y, batch_size = batch_size)

 

157/157 [==============================] - 5s 31ms/step - loss: 0.0444 - accuracy: 0.9860

Out[9]:

[0.044350724667310715, 0.9860000014305115]

 

결과 확인¶

 

Input으로 들어갈 이미지 데이터 확인

In [10]:

import matplotlib.pyplot as plt

import numpy as np

%matplotlib inline

In [11]:

test_image = test_x[0,:,:,0]
test_image.shape

Out[11]:

(28, 28)

In [12]:

plt.title(test_y[0])
plt.imshow(test_image,'gray')
plt.show()

 

 

• 모델에 Input Data로 확인 할 이미지 데이터 넣기

In [13]:

pred = model.predict(test_image.reshape(1,28,28,1))

In [14]:

pred.shape

Out[14]:

(1, 10)

In [16]:

pred  # softmax를 거친 각 레이블별 확률
      # 가장 높은 값을 가진 레이블이 예측된 정답

Out[16]:

array([[7.9267153e-09, 1.5796324e-07, 5.4426841e-06, 2.7217495e-05,
        1.0247860e-08, 4.9534464e-08, 3.9703126e-11, 9.9994802e-01,
        1.8639984e-08, 1.9035480e-05]], dtype=float32)

 

• np.argmax

In [17]:

np.argmax(pred)

Out[17]:

7

 

Test Batch¶

 

Batch로 Test Dataset 넣기

In [18]:

test_batch = test_x[:32]
test_batch.shape

Out[18]:

(32, 28, 28, 1)

 

Batch Test Dataset 모델에 넣기

In [20]:

preds = model.predict(test_batch)
preds.shape

Out[20]:

(32, 10)

 

• 결과 확인

In [23]:

np.argmax(preds).shape
np.argmax(preds,-1).shape
np.argmax(preds,-1)

Out[23]:

array([7, 2, 1, 0, 4, 1, 4, 9, 5, 9, 0, 6, 9, 0, 1, 5, 9, 7, 3, 4, 9, 6,
       6, 5, 4, 0, 7, 4, 0, 1, 3, 1], dtype=int64)

In [25]:

plt.imshow(test_batch[1,:,:,0],'gray')
plt.show()

 

In [ ]: