VGG16 Code Implementation with kaggle datasets

import numpy as np
import keras
from keras.models import Model
from keras.layers import Flatten, Dense, Input, Conv2D, MaxPooling2D
from tensorflow.keras.preprocessing.image import ImageDataGenerator #data Augmentation

def VGG16_Brief(classes=2): # classes = 감지할 클래스 수
    img_rows, img_cols = 224, 224
    img_channels = 3    
    img_dim = (img_rows, img_cols, img_channels) #차원..shape 정의..
    
    img_input = Input(shape=img_dim) #튜플.. 변경 x
    x = Conv2D(64,(3,3),padding='same',activation = 'relu')(img_input) #1층
    x = Conv2D(64,(3,3),padding='same',activation = 'relu')(x) #2층
    x = MaxPooling2D((2,2),strides=(2,2))(x) #2층 maxpool

    x = Conv2D(128,(3,3),padding='same',activation = 'relu')(x) #3층
    #x = Conv2D(128,(3,3),padding='same',activation = 'relu')(x) #4층
    x = MaxPooling2D((2,2),strides=(2,2))(x) #4층 maxpool

    x = Conv2D(256,(3,3),padding='same',activation = 'relu')(x) #5층
    #x = Conv2D(256,(3,3),padding='same',activation = 'relu')(x) #6층
    x = Conv2D(256,(3,3),padding='same',activation = 'relu')(x) #7층
    x = MaxPooling2D((2,2),strides=(2,2))(x) #7층 maxpool

    #x = Conv2D(512,(3,3),padding='same',activation = 'relu')(x) #8층
    #x = Conv2D(512,(3,3),padding='same',activation = 'relu')(x) #9층
    #x = Conv2D(512,(3,3),padding='same',activation = 'relu')(x) #10층
    #x = MaxPooling2D((2,2),strides=(2,2))(x) #10층 maxpool

    #x = Conv2D(512,(3,3),padding='same',activation = 'relu')(x) #11층
    #x = Conv2D(512,(3,3),padding='same',activation = 'relu')(x) #12층
    #x = Conv2D(512,(3,3),padding='same',activation = 'relu')(x) #13층
    #x = MaxPooling2D((2,2),strides=(2,2))(x) #13층 maxpool

    x = Flatten()(x)
    x = Dense(256, activation = 'relu')(x) #14층
    x = Dense(256, activation = 'relu')(x) #15층
    x = Dense(classes, activation = 'softmax')(x)
    
    model = Model(inputs=img_input, output = x)
    return model    

model.compile(optimizer=keras.optimizers.Adam(learning_rate=0.0003), 
          loss='categorical_crossentropy', metrics=['accuracy'])

model = VGG16_Brief(classes = 2)
model.summary()

Model: "model_2"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_2 (InputLayer)         (None, 224, 224, 3)       0         
_________________________________________________________________
conv2d_6 (Conv2D)            (None, 224, 224, 64)      1792      
_________________________________________________________________
conv2d_7 (Conv2D)            (None, 224, 224, 64)      36928     
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 112, 112, 64)      0         
_________________________________________________________________
conv2d_8 (Conv2D)            (None, 112, 112, 128)     73856     
_________________________________________________________________
max_pooling2d_5 (MaxPooling2 (None, 56, 56, 128)       0         
_________________________________________________________________
conv2d_9 (Conv2D)            (None, 56, 56, 256)       295168    
_________________________________________________________________
conv2d_10 (Conv2D)           (None, 56, 56, 256)       590080    
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 28, 28, 256)       0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 200704)            0         
_________________________________________________________________
dense_4 (Dense)              (None, 256)               51380480  
_________________________________________________________________
dense_5 (Dense)              (None, 256)               65792     
_________________________________________________________________
dense_6 (Dense)              (None, 2)                 514       
=================================================================
Total params: 52,444,610
Trainable params: 52,444,610
Non-trainable params: 0
_________________________________________________________________
C:\Users\notou\Anaconda3\envs\tf_gpu_10\lib\site-packages\ipykernel_launcher.py:35: UserWarning: Update your `Model` call to the Keras 2 API: `Model(inputs=Tensor("in..., outputs=Tensor("de...)`

training_dir='C://dataset//data_10000'
validation_dir='C://dataset//valid_1000_v2'


train_datagen = ImageDataGenerator(
rescale=1./255, #normalization 이미지 픽셀 값이 1/225를 곱해준다. default = 1
shear_range=0.2,#반시계 방향 밀리기
zoom_range=0.2, #줌
horizontal_flip=True #인풋을 무작위로 가로로 뒤집기
)

test_datagen = ImageDataGenerator(rescale = 1./255)

training_set = train_datagen.flow_from_directory( #폴더 형태로된 데이터 구조 바로 가저옴
                                                    #이 데이터 불러올 떄 앞에 정의한 파라미터로 전처리
training_dir,
target_size=(224,224), #폴더 내 이미지 사이즈 조정
batch_size=32,  #32장씩 읽어들임
#shuffle=True,
class_mode='categorical' #2D one-hot 인코딩된 라벨 반환
)

validation_set = test_datagen.flow_from_directory(
validation_dir,
target_size = (224, 224),
batch_size=32,
class_mode = 'categorical'    
)


#learnng rate 조정 필요할 시 조정해주는 code
from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau
learning_rate_reduction = ReduceLROnPlateau(monitor='val_accuracy', 
                                            patience=5, 
                                            verbose=1, 
                                            factor=0.1, 
                                            min_lr=0.00001)

history = model.fit_generator(training_set,
               steps_per_epoch=100, #에포크 한 번 돌떄, 데이터를 몇 번 볼것인가(training데이터수/배치사이즈)
               epochs = 100,
               validation_data= validation_set,
               callbacks=[learning_rate_reduction],
               validation_steps = 100 #한 번 에포크를 돌고난 후, validation set을 통해 validation 
                                     #accuracy를 측정할 때 validation set을 몇 번 볼것인가(valdation data/batch size)
               )

Epoch 1/100
100/100 [==============================] - 63s 628ms/step - loss: 0.7020 - accuracy: 0.4925 - val_loss: 0.6920 - val_accuracy: 0.5000
Epoch 2/100
100/100 [==============================] - 57s 568ms/step - loss: 0.6926 - accuracy: 0.5166 - val_loss: 0.6875 - val_accuracy: 0.4972

Epoch 50/100
100/100 [==============================] - 47s 474ms/step - loss: 0.2040 - accuracy: 0.9197 - val_loss: 0.2866 - val_accuracy: 0.8920

Epoch 100/100
100/100 [==============================] - 45s 448ms/step - loss: 0.1782 - accuracy: 0.9284 - val_loss: 0.1737 - val_accuracy: 0.8932

test_dir = 'C://dataset//test_1000_v2'
test_generator = test_datagen.flow_from_directory(
        test_dir,
        target_size=(224, 224),
        batch_size=32,
        class_mode='categorical')

test_loss, test_acc = model.evaluate_generator(test_generator, steps=50)
print('test loss:', test_loss)
print('test acc:', test_acc)

Found 2000 images belonging to 2 classes.
test loss: 0.4005548357963562
test acc: 0.9087499976158142


import matplotlib.pyplot as plt
plt.plot(history.history["accuracy"], 'y', label= 'ACCURACY')
plt.plot(history.history['loss'], 'r', label= 'LOSS')
plt.plot(history.history["val_loss"], 'b', label= 'VAL_LOSS')
plt.plot(history.history["val_accuracy"], 'g', label= 'VAL_ACC')
#plt.plot(test_acc, 'r', label= 'TEST_ACC')
plt.ylabel("Accuracy")
plt.xlabel("Epoch")
plt.legend(loc='lower right')
plt.show()