[케라스] Fashion-Mnist 기반 CNN모델 코드

from keras.datasets import mnist 
from keras.utils import np_utils
import numpy as np 
import sys 

데이터 셋 준비 

import keras 
fm = keras.datasets.fashion_mnist
(trainImage, trainLabel), (testImage, testLabel)= fm.load_data()

xTrain=trainImage.reshape(xTrain.shape[0],28,28,1)
xTrain=xTrain.astype("float64")
xTrain=xTrain/255

xTest=testImage.reshape(xTest.shape[0],28,28,1).astype("float64")/255

yTrain=np_utils.to_categorical(trainLabel,10)
yTest=np_utils.to_categorical(testLabel,10)

print(xTrain.shape)
print(yTrain.shape)
print(xTest.shape)
print(yTest.shape)

>>>
(60000, 28, 28, 1)
(60000, 10)
(10000, 28, 28, 1)
(10000, 10)

 

import matplotlib.pyplot as plt

plt.imshow(xTrain[:,:,:,0][0], cmap="Greys")
plt.show()

라이브러리 호출 

from keras.models import Sequential
from keras.layers import Dense 
from keras.callbacks import ModelCheckpoint, EarlyStopping
import numpy as np 
import tensorflow as tf 
import os
from keras.layers import *

모델 생성 

# 데이터 형상 관련 상수 정의 
IMAGE_WIDTH=28
IMAGE_HEIGHT=28
IMAGE_SIZE=(IMAGE_WIDTH, IMAGE_HEIGHT)
IMAGE_CHANNEL=1

# 레이어 1
model = Sequential()
model.add(Conv2D(32, kernel_size=2, activation="relu", input_shape=(IMAGE_HEIGHT, IMAGE_WIDTH , IMAGE_CHANNEL)))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.25))

# 레이어 2 
model.add(Conv2D(64,kernel_size=2, activation="relu"))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.25))

# 레이어3
model.add(Conv2D(128, kernel_size=2,activation="relu"))
model.add(BatchNormalization())
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.25))
          
# Fully Connected 
model.add(Flatten())
model.add(Dense(512, activation='relu'))
model.add(BatchNormalization())
model.add(Dropout(0.5))

model.add(Dense(10,activation="softmax"))

모델 환경설정 

# 모델 실행 옵션 
model.compile(loss="categorical_crossentropy", optimizer="rmsprop", metrics=['accuracy'])
model.summary()
>>>
Model: "sequential_10"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_18 (Conv2D)           (None, 27, 27, 32)        160       
_________________________________________________________________
batch_normalization_20 (Batc (None, 27, 27, 32)        128       
_________________________________________________________________
max_pooling2d_17 (MaxPooling (None, 13, 13, 32)        0         
_________________________________________________________________
dropout_22 (Dropout)         (None, 13, 13, 32)        0         
_________________________________________________________________
conv2d_19 (Conv2D)           (None, 12, 12, 64)        8256      
_________________________________________________________________
batch_normalization_21 (Batc (None, 12, 12, 64)        256       
_________________________________________________________________
max_pooling2d_18 (MaxPooling (None, 6, 6, 64)          0         
_________________________________________________________________
dropout_23 (Dropout)         (None, 6, 6, 64)          0         
_________________________________________________________________
conv2d_20 (Conv2D)           (None, 5, 5, 128)         32896     
_________________________________________________________________
batch_normalization_22 (Batc (None, 5, 5, 128)         512       
_________________________________________________________________
max_pooling2d_19 (MaxPooling (None, 2, 2, 128)         0         
_________________________________________________________________
dropout_24 (Dropout)         (None, 2, 2, 128)         0         
_________________________________________________________________
flatten_7 (Flatten)          (None, 512)               0         
_________________________________________________________________
dense_11 (Dense)             (None, 512)               262656    
_________________________________________________________________
batch_normalization_23 (Batc (None, 512)               2048      
_________________________________________________________________
dropout_25 (Dropout)         (None, 512)               0         
_________________________________________________________________
dense_12 (Dense)             (None, 10)                5130      
=================================================================
Total params: 312,042
Trainable params: 310,570
Non-trainable params: 1,472
_________________________________________________________________

모델 실행 

# reduceLROnPlateau 
# : callback 함수의 일종, learning rate가 더이상 업데이트가 되지 않으면, 학습을 중단하여라 
from keras.callbacks import EarlyStopping, ReduceLROnPlateau

earlystop = EarlyStopping(patience=10)
learning_rate_reduction=ReduceLROnPlateau(
                        monitor= "val_accuracy", 
                        patience = 2, 
                        factor = 0.5, 
                        min_lr=0.0001,
                        verbose=1)
                        
callbacks = [earlystop, learning_rate_reduction]

#모델 학습

history= model.fit(xTrain, yTrain,validation_data=(xTest,yTest),
                   epochs=30, batch_size=200, callbacks=callbacks)

 

historyDict=history.history

acc=history.history['accuracy']
val_acc=history.history['val_accuracy']
loss=history.history['loss']
val_loss=history.history['val_loss']

%matplotlib notebook
epo = range(1, len(acc)+1)
plt.plot(epo, loss, 'bo', label="Traing loss")
plt.plot(epo, val_loss, 'b', label="Val loss")
plt.xlabel("epoch")
plt.ylabel("Loss")
plt.legend()
plt.show()

%matplotlib notebook

plt.plot(epo, acc, 'ro', label="Traing accuracy")
plt.plot(epo, val_acc, 'r', label="Val accuracy")
plt.xlabel("epoch")
plt.ylabel("Accuracy")
plt.legend()
plt.show()

모델 평가

# 모델평가
score = model.evaluate(xTest, yTest, verbose=0)
print('\n', '테스트 정확도:', score[1])

>>> 테스트 정확도: 0.901199996471405