tf_CNN

이미지 처리입니다. 

합성곱에 대한 간단한 코드입니다.

CNN에 관해 설명이 잘 되어 있는 블로그가 있어 링크 남깁니다.

(http://taewan.kim/post/cnn/)

 

 

<참조 : https://www.tensorflow.org/tutorials/images/cnn>

In [ ]:

Convolution Neural Network(CNN)¶

• cnn에 관한 설명은 아주 많습니다. 그중에 괜찮은 블로그가 있어 안내합니다.
  • http://taewan.kim/post/cnn/

01. Import¶

In [2]:

import tensorflow as tf

from tensorflow.keras import datasets, layers, models
import matplotlib.pyplot as plt

In [ ]:

02. Download and prepare the CIFAR10 dataset¶

In [3]:

(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()

# Normalize pixel values to be between 0 and 1
train_images, test_images = train_images / 255.0, test_images / 255.0

Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz
170500096/170498071 [==============================] - 100s 1us/step
170508288/170498071 [==============================] - 100s 1us/step

In [ ]:

03. Verify the data¶

In [4]:

class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer',
              'dog', 'frog', 'horse', 'ship', 'truck']

plt.figure(figsize=(10, 10))
for i in range(25):
    plt.subplot(5, 5, i+1)
    plt.xticks([])
    plt.yticks([])
    plt.grid(False)
    plt.imshow(train_images[i])
    
    # The CIFAR labels happen to be arrays,
    # which is why you need the extra index
    plt.xlabel(class_names[train_labels[i][0]])
    
plt.show()

In [ ]:

03. Create the convolutional base¶

In [6]:

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))

In [7]:

model.summary()

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 30, 30, 32)        896       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 13, 13, 64)        18496     
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 6, 6, 64)          0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 4, 4, 64)          36928     
=================================================================
Total params: 56,320
Trainable params: 56,320
Non-trainable params: 0
_________________________________________________________________

In [ ]:

Add Dense layers on top¶

• CIFAR has 10 output classes.

In [8]:

model.add(layers.Flatten())
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(10))

In [9]:

model.summary()

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 30, 30, 32)        896       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 15, 15, 32)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 13, 13, 64)        18496     
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 6, 6, 64)          0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 4, 4, 64)          36928     
_________________________________________________________________
flatten (Flatten)            (None, 1024)              0         
_________________________________________________________________
dense (Dense)                (None, 64)                65600     
_________________________________________________________________
dense_1 (Dense)              (None, 10)                650       
=================================================================
Total params: 122,570
Trainable params: 122,570
Non-trainable params: 0
_________________________________________________________________

In [ ]:

04. Compile and train the model¶

In [10]:

model.compile(optimizer='adam',
             loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
             metrics=['accuracy'])

In [11]:

history = model.fit(train_images, train_labels, epochs=10,
                   validation_data=(test_images, test_labels))

Epoch 1/10
1563/1563 [==============================] - 216s 132ms/step - loss: 1.5220 - accuracy: 0.4445 - val_loss: 1.3215 - val_accuracy: 0.5190
Epoch 2/10
1563/1563 [==============================] - 214s 137ms/step - loss: 1.1722 - accuracy: 0.5841 - val_loss: 1.0847 - val_accuracy: 0.6145
Epoch 3/10
1563/1563 [==============================] - 191s 122ms/step - loss: 1.0172 - accuracy: 0.6440 - val_loss: 0.9958 - val_accuracy: 0.6519
Epoch 4/10
1563/1563 [==============================] - 186s 119ms/step - loss: 0.9163 - accuracy: 0.6811 - val_loss: 0.9272 - val_accuracy: 0.6783
Epoch 5/10
1563/1563 [==============================] - 201s 129ms/step - loss: 0.8489 - accuracy: 0.7037 - val_loss: 0.8974 - val_accuracy: 0.6879
Epoch 6/10
1563/1563 [==============================] - 207s 132ms/step - loss: 0.7922 - accuracy: 0.7232 - val_loss: 0.9083 - val_accuracy: 0.6805
Epoch 7/10
1563/1563 [==============================] - 206s 132ms/step - loss: 0.7467 - accuracy: 0.7391 - val_loss: 0.9076 - val_accuracy: 0.6856
Epoch 8/10
1563/1563 [==============================] - 206s 132ms/step - loss: 0.7057 - accuracy: 0.7524 - val_loss: 0.8566 - val_accuracy: 0.7060
Epoch 9/10
1563/1563 [==============================] - 206s 132ms/step - loss: 0.6717 - accuracy: 0.7640 - val_loss: 0.8637 - val_accuracy: 0.7071
Epoch 10/10
1563/1563 [==============================] - 207s 132ms/step - loss: 0.6376 - accuracy: 0.7751 - val_loss: 0.8780 - val_accuracy: 0.6987

In [ ]:

05. Evaluate the model¶

In [12]:

plt.plot(history.history['accuracy'], label='accuracy')
plt.plot(history.history['val_accuracy'], label='val_accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.ylim([0.5, 1])
plt.legend(loc='lower right')

test_loss, test_acc = model.evaluate(test_images, test_labels, verbose=2)

313/313 - 13s - loss: 0.8780 - accuracy: 0.6987

In [14]:

plt.plot(history.history['loss'], label='loss')
plt.plot(history.history['val_loss'], label='val_loss')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.ylim([0, 2])
plt.legend(loc='upper right')

Out[14]:

<matplotlib.legend.Legend at 0xb139521700>

In [ ]: