인공지능 경진대회 - Kaggle Cat&Dog Dataset 이진분류

라이브러리 import¶

In [72]:

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from tensorflow.keras.preprocessing.image import ImageDataGenerator,img_to_array,load_img

내가 만든 데이터셋 로딩 및 전처리¶

-  데이터셋 주소:  https://www.kaggle.com/shaunthesheep/microsoft-catsvsdogs-dataset

In [2]:

import shutil
import os

# 입력한 경로에 폴더를 생성하는 메소드
def make_dir(target_dir): # 입력값 :처리 대상 폴더 경로 
    try:
        if not os.path.exists(target_dir): # 경로 내에서 설명 이름으로 된 폴더가 없다면 
            os.makedirs(target_dir) #설명 이름으로 된 폴더 생성 

    except OSError:
        print ('Error: Creating directory. ' + target_dir)

# data를 from_path에서 target_path로 이동시키는 메소드
def transfer(from_path,target_path):
        shutil.move(from_path, target_path)

def preprocessing_datasets(classes,from_path, target_path): # 클래스 (*리스트), 대상 폴더, 파일(클래스)이름, 타겟 폴더 
    # 학습, 검증, test용 경로 생성 
    datasets_dir = [os.path.join(target_path,"train"), os.path.join(target_path,"val"), os.path.join(target_path,"test")]
    for path in datasets_dir:
        make_dir(path)
    # 
    for class_name in classes: # 각 클래스명마다 
        name = os.listdir(from_path) # 데이터셋 지정 경로 받아오기
        class_data =  [os.path.join(from_path,i) for i in name if "{}".format(class_name) in i] # 데이터셋에서 해당 클래스 네임을 포함한 경로들을 절대경로로 받아옴
        print("class({class_name})_num:",len(class_data))
        val_num = int(len(class_data)*0.8) # 해당 클래스 데이터 갯수중 80%는 학습용, 20%는 검증용
        train_data = class_data[ : val_num] # 맨앞~ 80%를 학습용으로 지정
        print("trian_num:",len(train_data))
        val_data = class_data[val_num : ] # 80% ~ 맨 뒤 순서까지 검증용 지정
        print("val_num:",len(val_data))

        train_data_dir = os.path.join(datasets_dir[0], class_name)
        make_dir(train_data_dir) # train forder에 해당 클래스명 폴더 생성 // keras preprocessing 으로 데이터셋 정형화 시킬 거임
        val_data_dir = os.path.join(datasets_dir[1], class_name) 
        make_dir(val_data_dir)   # val forder에 해당 클래스명 폴더 생성
        for train in train_data: #  옮기기 
            if len(os.listdir(train_data_dir))< len(train_data):
                transfer(train, train_data_dir)   
                
        for val in val_data:
            if len(os.listdir(val_data_dir))<len(val_data):
                transfer(val, val_data_dir)
        # 다 옮겨졌으면 validation set에서 test dataset 분할
        val_list = os.listdir(val_data_dir) # val폴더 내 class명 가진 폴더 안에 있는거 다 가져오기
        test_list = [os.path.join(val_data_dir,i) for i in val_list if "{}".format(class_name) in i] #절대경로는 val_data_dir/해당파일
        test_num = int(len(test_list)*0.2) # 그 개수중 20%를  테스트용으로 쓸 것이다 
        print("test_num:",test_num)
        test_data = test_list[ : test_num] # 0~20%까지 지정
        test_data_dir = os.path.join(datasets_dir[2],class_name) # 테스트셋 폴더 경로 지정 및 생성
        make_dir(test_data_dir)
        for test in test_data:
            if len(os.listdir(test_data_dir)) < test_num: # 테스트셋으로 분리를 하고 났으면 파일 개수가 test_num과 같아지므로 중복분리 방지  
                transfer(test, test_data_dir)



# 학습 된 모델의 학습 과정 시각화
def plot_model__hist(hist):
    plt.figure(figsize=(6,6))
    plt.style.use("ggplot")
    plt.plot(hist.history['loss'], color='b', label="Training loss")
    plt.plot(hist.history['val_loss'], color='r', label="Validation loss")
    plt.legend()
    plt.show()

    plt.figure()

    plt.figure(figsize=(6,6))
    plt.style.use("ggplot")
    plt.plot(hist.history['accuracy'], color='b', label="Training accuracy")
    plt.plot(hist.history['val_accuracy'], color='r',label="Validation accuracy")
    plt.legend(loc = "lower right")
    plt.show()

from_path = r"C:\Users\cvLab\Desktop\from_dataset" #데이터셋 받아져있는 경로
# from_path = r"C:\Users\cvLab\Downloads\archive(1)\train\train" #데이터셋 받아져있는 경로
target_path = r"C:\Users\cvLab\Desktop\dataset" # 데이터셋을 생성할 경로
# target_path = r"D:\Project(2)\cat&dog_set" # 데이터셋을 생성할 경로

classes = ['cat', 'dog']
# preprocessing_datasets(classes,from_path,target_path)

Train, test, validation 데이터셋 텐서화 시켜서 학습 준비 하기¶

In [3]:

#데이터셋 리턴하는 메소드 
def get_dataset(dataset_path=r"D:\Project(2)\cat&dog_set",size=128, aug=False, crawl=False,batch_size=32,kfold=False): #이미지 어그먼테이션, 크롤링된 데이터셋도 포함할지 선택
    if aug :
        train_datagen = ImageDataGenerator(
        rescale=1./255,
        rotation_range = 30,
        shear_range = 0.2,
        zoom_range = 0.2,
        vertical_flip = True)
    else:     
        train_datagen = ImageDataGenerator(rescale=1./255)
    val_datagen = ImageDataGenerator(rescale=1/255)
    test_datagen = ImageDataGenerator(rescale=1/255)

    if crawl:
        train_path = os.path.join(dataset_path,"additional_train")
    if kfold:
        train_path = os.path.join(dataset_path,"k-ford_train")

    else:
        train_path = os.path.join(dataset_path,"train")
    train_set = train_datagen.flow_from_directory(directory=train_path,target_size=(size,size),batch_size=batch_size,color_mode="rgb",class_mode='binary')
    
    val_path = os.path.join(dataset_path,"val")
    val_set = val_datagen.flow_from_directory(directory=val_path,target_size=(size,size),batch_size=int(batch_size*0.8),color_mode="rgb",class_mode='binary')
    
    test_path = os.path.join(dataset_path,"test")
    test_set = test_datagen.flow_from_directory(directory=test_path,target_size=(size,size),batch_size=10,color_mode="rgb",class_mode='binary')

    return train_set, val_set, test_set
    
train_set, val_set, test_set = get_dataset(size=128)

Found 20000 images belonging to 2 classes.
Found 4000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.

임의로 신경망 생성¶

• Conv1 = 5x5x64
• Max Fooling = 2x2 stride = 2
• Conv2 = 5x5x32
• Max Fooling = 2x2 stride = 2
• Conv2 = 5x5x16
• Max Fooling = 2x2 stride = 2
• FC1 = 256 (relu)
• FC2 = 1 (sigmoid)

In [5]:

from tensorflow.keras import Model
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D,Dense,MaxPooling2D,Flatten,Dropout
model = Sequential([
    Conv2D(filters = 64, kernel_size = (5,5), activation = "relu", input_shape = (128,128,3)),
    MaxPooling2D(pool_size = (2,2)),
    Conv2D(filters = 32, kernel_size = (5,5), activation = "relu"),
    MaxPooling2D(pool_size = (2,2)),
    Dropout(0.25),
    Conv2D(filters = 16, kernel_size = (5,5), activation = "relu"),
    MaxPooling2D(pool_size = (2,2)),
    Dropout(0.25),
    Flatten(),
    Dense(units = 256, activation = "relu"),
    Dropout(0.5),
    Dense(units = 1, activation = "sigmoid")
])

In [14]:

# rmsprop 기법으로 최적화, logistic regression을 위해 binary CrossEntropy loss를 적용 froma
from tensorflow.keras.optimizers import RMSprop
opt = RMSprop()
model.compile(optimizer=opt,loss='binary_crossentropy',metrics=['accuracy']) 

10 epochs 만 학습¶

In [ ]:

hist = model.fit(train_set, validation_data = val_set, epochs = 10)

Epoch 1/10
625/625 [==============================] - 138s 220ms/step - loss: 0.6942 - accuracy: 0.5321 - val_loss: 0.6687 - val_accuracy: 0.5840
Epoch 2/10
625/625 [==============================] - 32s 51ms/step - loss: 0.6693 - accuracy: 0.5929 - val_loss: 0.6449 - val_accuracy: 0.6350
Epoch 3/10
625/625 [==============================] - 32s 51ms/step - loss: 0.6232 - accuracy: 0.6622 - val_loss: 0.5713 - val_accuracy: 0.7105
Epoch 4/10
625/625 [==============================] - 32s 51ms/step - loss: 0.5853 - accuracy: 0.6985 - val_loss: 0.5488 - val_accuracy: 0.7265
Epoch 5/10
625/625 [==============================] - 32s 51ms/step - loss: 0.5455 - accuracy: 0.7311 - val_loss: 0.5487 - val_accuracy: 0.7418
Epoch 6/10
625/625 [==============================] - 32s 50ms/step - loss: 0.5101 - accuracy: 0.7532 - val_loss: 0.4859 - val_accuracy: 0.7623
Epoch 7/10
625/625 [==============================] - 32s 50ms/step - loss: 0.4745 - accuracy: 0.7783 - val_loss: 0.4606 - val_accuracy: 0.7918
Epoch 8/10
625/625 [==============================] - 32s 50ms/step - loss: 0.4517 - accuracy: 0.7929 - val_loss: 0.4182 - val_accuracy: 0.8045
Epoch 9/10
625/625 [==============================] - 32s 50ms/step - loss: 0.4289 - accuracy: 0.8081 - val_loss: 0.4358 - val_accuracy: 0.7993
Epoch 10/10
625/625 [==============================] - 32s 51ms/step - loss: 0.4096 - accuracy: 0.8196 - val_loss: 0.3833 - val_accuracy: 0.8355

In [ ]:

plot_model__hist(hist)
loss,acc = model.evaluate(test_set, verbose=2)
print("10 epochs 학습 후 모델의 정확도: {:5.2f}%".format(100*acc))
print("10 epochs 학습 후 모델의 Loss: {}".format(loss))

<Figure size 432x288 with 0 Axes>

100/100 - 1s - loss: 0.3695 - accuracy: 0.8510
10 epochs 학습 후 모델의 정확도: 85.10%
10 epochs 학습 후 모델의 Loss: 0.36947640776634216

1. ResNet50V2 전이학습을 이용해서 최고의 성능을 이끌어보기¶

In [19]:

from tensorflow import keras
from tensorflow.keras import Model
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D,Dense,MaxPooling2D,Flatten,Dropout
# from tensorflow.python.keras.applications.resnet import ResNet50
from tensorflow.python.keras.applications.resnet_v2 import ResNet50V2
from tensorflow.keras.callbacks import EarlyStopping , ReduceLROnPlateau
from tensorflow.keras.layers import GlobalAveragePooling2D
import tensorflow as tf


def get_ResNet50V2(size=128,trainable=False):
    base_model = ResNet50V2(input_shape=(size,size,3), include_top=False, weights='imagenet') # 백본 설정
    base_model.trainable = trainable # 백본 학습 불가능하도록 freezing
    global_average_layer = GlobalAveragePooling2D() # 전체 채널을 평균내어 이동하며 학습 

    ResNet50V2model = tf.keras.Sequential([ #back born + global avg Pooling2D 층 + FC 설정 
        base_model,
        global_average_layer,
        Flatten()  
        ])
    return ResNet50V2model

def get_callbacks(checkpoint_path, patience = 5):
    
    earlystop = EarlyStopping(
    monitor='val_accuracy', min_delta=0.0001,
    patience=patience)
  
    cp_callback = tf.keras.callbacks.ModelCheckpoint(filepath=checkpoint_path,
                                                 save_weights_only=True,
                                                 verbose=1)
  
    callbacks = [earlystop, cp_callback]
    return callbacks

1차 신경망 구현¶

• FC계층 구성
• callback함수 정의
• Back Born 프리징 후 FC 계층만 1차학습
• optimizer 별로 성능 확인 후 Back Born 프리징 해제 후 재학습

1. optimizer : Adam¶

FC1 :256, relu (dropout 50%)¶

FC2 :128, relu (dropout 25%)¶

• 결과 = 95.1%

In [16]:

# adam으로 해보기
ResNet50V2_model_1 = Sequential([
    get_ResNet50V2(size=128),
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])

checkpoint_path_1 = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path_1)

from tensorflow.keras.optimizers import Adam
opt = Adam(learning_rate=0.001)
ResNet50V2_model_1.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

In [14]:

ResNet50V2_model_hist_1 = ResNet50V2_model_1.fit(train_set,
                    validation_data=val_set,
                    epochs=100,
                    callbacks = [callbacks])

Epoch 1/100
625/625 [==============================] - 138s 201ms/step - loss: 0.2764 - accuracy: 0.8995 - val_loss: 0.1285 - val_accuracy: 0.9490

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0001.ckpt
Epoch 2/100
625/625 [==============================] - 34s 55ms/step - loss: 0.1411 - accuracy: 0.9448 - val_loss: 0.1271 - val_accuracy: 0.9488

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0002.ckpt
Epoch 3/100
625/625 [==============================] - 35s 56ms/step - loss: 0.1258 - accuracy: 0.9491 - val_loss: 0.1181 - val_accuracy: 0.9510

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0003.ckpt
Epoch 4/100
625/625 [==============================] - 35s 56ms/step - loss: 0.1172 - accuracy: 0.9561 - val_loss: 0.1191 - val_accuracy: 0.9470

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0004.ckpt
Epoch 5/100
625/625 [==============================] - 36s 57ms/step - loss: 0.1079 - accuracy: 0.9586 - val_loss: 0.1142 - val_accuracy: 0.9532

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0005.ckpt
Epoch 6/100
625/625 [==============================] - 35s 56ms/step - loss: 0.0947 - accuracy: 0.9622 - val_loss: 0.1146 - val_accuracy: 0.9563

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0006.ckpt
Epoch 7/100
625/625 [==============================] - 35s 56ms/step - loss: 0.0895 - accuracy: 0.9661 - val_loss: 0.1175 - val_accuracy: 0.9548

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0007.ckpt
Epoch 8/100
625/625 [==============================] - 35s 55ms/step - loss: 0.0905 - accuracy: 0.9651 - val_loss: 0.1201 - val_accuracy: 0.9545

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0008.ckpt
Epoch 9/100
625/625 [==============================] - 37s 59ms/step - loss: 0.0761 - accuracy: 0.9711 - val_loss: 0.1183 - val_accuracy: 0.9557

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0009.ckpt

In [16]:

plot_model__hist(ResNet50V2_model_hist_1)
loss_1,acc_1 = ResNet50V2_model_1.evaluate(test_set, verbose=2)
print("ResNet50V2_model_1 백본 프리징 후 모델의 정확도: {:5.2f}%".format(100*acc_1))
print("ResNet50V2_model_1 백본 프리징 후 모델의 Loss: {}".format(loss_1))

<Figure size 432x288 with 0 Axes>

100/100 - 6s - loss: 0.1386 - accuracy: 0.9510
ResNet50V2_model_1 백본 프리징 후 모델의 정확도: 95.10%
ResNet50V2_model_1 백본 프리징 후 모델의 Loss: 0.138584166765213

2. optimizer : SGD¶

FC1 :256, relu (dropout 50%)¶

FC2 :128, relu (dropout 25%)¶

• 결과 = 94.5%

In [7]:

# SGD로 해보기
ResNet50V2_model_1 = keras.Sequential([
    get_ResNet50V2(size=128),
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])

checkpoint_path_1 = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path_1)

from tensorflow.keras.optimizers import Adam, SGD
opt = SGD(lr=0.01, decay=1e-6, momentum=0.001,nesterov=True)
ResNet50V2_model_1.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])
ResNet50V2_model_hist_1 = ResNet50V2_model_1.fit(train_set,
                    validation_data=val_set,
                    epochs=100,
                    callbacks = [callbacks])

In [27]:

plot_model__hist(ResNet50V2_model_hist_1)
loss_1,acc_1 = ResNet50V2_model_1.evaluate(test_set, verbose=2)
print("ResNet50V2_model_1 백본 프리징 후 모델의 정확도: {:5.2f}%".format(100*acc_1))
print("ResNet50V2_model_1 백본 프리징 후 모델의 Loss: {}".format(loss_1))

<Figure size 432x288 with 0 Axes>

100/100 - 2s - loss: 0.1326 - accuracy: 0.9450
ResNet50V2_model_1 백본 프리징 후 모델의 정확도: 94.50%
ResNet50V2_model_1 백본 프리징 후 모델의 Loss: 0.1326497346162796

3. optimizer : Adadelta¶

FC1 :256, relu (dropout 50%)¶

FC2 :128, relu (dropout 25%)¶

• 결과 = 92.7%

In [28]:

# Adadelta로 해보기
ResNet50V2_model_1 = keras.Sequential([
    get_ResNet50V2(size=128),
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])

checkpoint_path_1 = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path_1)

from tensorflow.keras.optimizers import Adam, SGD, Adadelta
opt = Adadelta(rho=0.95)
ResNet50V2_model_1.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])
ResNet50V2_model_hist_1 = ResNet50V2_model_1.fit(train_set,
                    validation_data=val_set,
                    epochs=100,
                    callbacks = [callbacks])

Epoch 1/100
625/625 [==============================] - 38s 57ms/step - loss: 1.3201 - accuracy: 0.5187 - val_loss: 0.5449 - val_accuracy: 0.7308

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0001.ckpt
Epoch 2/100
625/625 [==============================] - 34s 54ms/step - loss: 0.9460 - accuracy: 0.6148 - val_loss: 0.4102 - val_accuracy: 0.8152

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0002.ckpt
Epoch 3/100
625/625 [==============================] - 34s 54ms/step - loss: 0.7657 - accuracy: 0.6805 - val_loss: 0.3356 - val_accuracy: 0.8530

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0003.ckpt
Epoch 4/100
625/625 [==============================] - 34s 54ms/step - loss: 0.6358 - accuracy: 0.7278 - val_loss: 0.2884 - val_accuracy: 0.8758

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0004.ckpt
Epoch 5/100
625/625 [==============================] - 33s 52ms/step - loss: 0.5586 - accuracy: 0.7632 - val_loss: 0.2576 - val_accuracy: 0.8895

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0005.ckpt
Epoch 6/100
625/625 [==============================] - 33s 53ms/step - loss: 0.5035 - accuracy: 0.7899 - val_loss: 0.2361 - val_accuracy: 0.9010

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0006.ckpt
Epoch 7/100
625/625 [==============================] - 33s 53ms/step - loss: 0.4554 - accuracy: 0.8061 - val_loss: 0.2208 - val_accuracy: 0.9060

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0007.ckpt
Epoch 8/100
625/625 [==============================] - 33s 52ms/step - loss: 0.4379 - accuracy: 0.8194 - val_loss: 0.2095 - val_accuracy: 0.9110

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0008.ckpt
Epoch 9/100
625/625 [==============================] - 33s 52ms/step - loss: 0.4106 - accuracy: 0.8257 - val_loss: 0.2009 - val_accuracy: 0.9160

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0009.ckpt
Epoch 10/100
625/625 [==============================] - 33s 53ms/step - loss: 0.3891 - accuracy: 0.8427 - val_loss: 0.1937 - val_accuracy: 0.9187

Epoch 00010: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0010.ckpt
Epoch 11/100
625/625 [==============================] - 35s 56ms/step - loss: 0.3543 - accuracy: 0.8524 - val_loss: 0.1883 - val_accuracy: 0.9197

Epoch 00011: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0011.ckpt
Epoch 12/100
625/625 [==============================] - 36s 58ms/step - loss: 0.3569 - accuracy: 0.8513 - val_loss: 0.1836 - val_accuracy: 0.9205

Epoch 00012: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0012.ckpt
Epoch 13/100
625/625 [==============================] - 35s 56ms/step - loss: 0.3457 - accuracy: 0.8584 - val_loss: 0.1798 - val_accuracy: 0.9200

Epoch 00013: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0013.ckpt
Epoch 14/100
625/625 [==============================] - 33s 53ms/step - loss: 0.3363 - accuracy: 0.8616 - val_loss: 0.1771 - val_accuracy: 0.9212

Epoch 00014: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0014.ckpt
Epoch 15/100
625/625 [==============================] - 33s 53ms/step - loss: 0.3261 - accuracy: 0.8705 - val_loss: 0.1742 - val_accuracy: 0.9225

Epoch 00015: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0015.ckpt
Epoch 16/100
625/625 [==============================] - 33s 52ms/step - loss: 0.3186 - accuracy: 0.8702 - val_loss: 0.1717 - val_accuracy: 0.9235

Epoch 00016: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0016.ckpt
Epoch 17/100
625/625 [==============================] - 33s 52ms/step - loss: 0.3046 - accuracy: 0.8736 - val_loss: 0.1693 - val_accuracy: 0.9243

Epoch 00017: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0017.ckpt
Epoch 18/100
625/625 [==============================] - 33s 53ms/step - loss: 0.2908 - accuracy: 0.8800 - val_loss: 0.1675 - val_accuracy: 0.9258

Epoch 00018: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0018.ckpt
Epoch 19/100
625/625 [==============================] - 39s 62ms/step - loss: 0.2917 - accuracy: 0.8808 - val_loss: 0.1659 - val_accuracy: 0.9280

Epoch 00019: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0019.ckpt
Epoch 20/100
625/625 [==============================] - 34s 55ms/step - loss: 0.2935 - accuracy: 0.8795 - val_loss: 0.1644 - val_accuracy: 0.9300

Epoch 00020: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0020.ckpt
Epoch 21/100
625/625 [==============================] - 33s 53ms/step - loss: 0.2887 - accuracy: 0.8839 - val_loss: 0.1630 - val_accuracy: 0.9298

Epoch 00021: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0021.ckpt
Epoch 22/100
625/625 [==============================] - 33s 53ms/step - loss: 0.2839 - accuracy: 0.8817 - val_loss: 0.1617 - val_accuracy: 0.9302

Epoch 00022: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0022.ckpt
Epoch 23/100
625/625 [==============================] - 33s 53ms/step - loss: 0.2843 - accuracy: 0.8884 - val_loss: 0.1604 - val_accuracy: 0.9305

Epoch 00023: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0023.ckpt
Epoch 24/100
625/625 [==============================] - 33s 53ms/step - loss: 0.2698 - accuracy: 0.8910 - val_loss: 0.1594 - val_accuracy: 0.9295

Epoch 00024: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0024.ckpt
Epoch 25/100
625/625 [==============================] - 33s 53ms/step - loss: 0.2755 - accuracy: 0.8899 - val_loss: 0.1585 - val_accuracy: 0.9302

Epoch 00025: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0025.ckpt
Epoch 26/100
625/625 [==============================] - 34s 54ms/step - loss: 0.2663 - accuracy: 0.8908 - val_loss: 0.1577 - val_accuracy: 0.9310

Epoch 00026: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0026.ckpt
Epoch 27/100
625/625 [==============================] - 34s 54ms/step - loss: 0.2617 - accuracy: 0.8931 - val_loss: 0.1565 - val_accuracy: 0.9317

Epoch 00027: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0027.ckpt
Epoch 28/100
625/625 [==============================] - 33s 52ms/step - loss: 0.2679 - accuracy: 0.8914 - val_loss: 0.1557 - val_accuracy: 0.9317

Epoch 00028: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0028.ckpt
Epoch 29/100
625/625 [==============================] - 33s 53ms/step - loss: 0.2637 - accuracy: 0.8914 - val_loss: 0.1549 - val_accuracy: 0.9333

Epoch 00029: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0029.ckpt
Epoch 30/100
625/625 [==============================] - 35s 55ms/step - loss: 0.2468 - accuracy: 0.9009 - val_loss: 0.1541 - val_accuracy: 0.9335

Epoch 00030: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0030.ckpt
Epoch 31/100
625/625 [==============================] - 34s 54ms/step - loss: 0.2569 - accuracy: 0.8963 - val_loss: 0.1534 - val_accuracy: 0.9337

Epoch 00031: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0031.ckpt
Epoch 32/100
625/625 [==============================] - 33s 52ms/step - loss: 0.2467 - accuracy: 0.8990 - val_loss: 0.1528 - val_accuracy: 0.9333

Epoch 00032: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0032.ckpt
Epoch 33/100
625/625 [==============================] - 31s 50ms/step - loss: 0.2544 - accuracy: 0.8979 - val_loss: 0.1521 - val_accuracy: 0.9335

Epoch 00033: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0033.ckpt
Epoch 34/100
625/625 [==============================] - 32s 50ms/step - loss: 0.2508 - accuracy: 0.8992 - val_loss: 0.1516 - val_accuracy: 0.9337

Epoch 00034: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0034.ckpt

In [29]:

plot_model__hist(ResNet50V2_model_hist_1)
loss_1,acc_1 = ResNet50V2_model_1.evaluate(test_set, verbose=2)
print("ResNet50V2_model_1 백본 프리징 후 모델의 정확도: {:5.2f}%".format(100*acc_1))
print("ResNet50V2_model_1 백본 프리징 후 모델의 Loss: {}".format(loss_1))

<Figure size 432x288 with 0 Axes>

100/100 - 2s - loss: 0.1535 - accuracy: 0.9270
ResNet50V2_model_1 백본 프리징 후 모델의 정확도: 92.70%
ResNet50V2_model_1 백본 프리징 후 모델의 Loss: 0.15354347229003906

<결과>

Adam = 95.1%¶

SGD = 94.5%¶

Adadelta = 92.7%¶

optimizer = Adam으로 채택하고 Back born Freezing 해제후 전체 신경망 학습¶

In [34]:

ResNet50V2_model = keras.Sequential([
    get_ResNet50V2(size=128,trainable=True), # 백본 학습 가능하게 설정
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])

checkpoint_path_1 = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path_1)

from tensorflow.keras.optimizers import Adam
opt = Adam(learning_rate=0.001)
ResNet50V2_model.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

In [35]:

ResNet50V2_model_hist = ResNet50V2_model.fit(train_set,
                    validation_data=val_set,
                    epochs=100,
                    callbacks = [callbacks])

Epoch 1/100
625/625 [==============================] - 86s 127ms/step - loss: 0.5450 - accuracy: 0.7480 - val_loss: 0.9595 - val_accuracy: 0.6102

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0001.ckpt
Epoch 2/100
625/625 [==============================] - 79s 126ms/step - loss: 0.2457 - accuracy: 0.8989 - val_loss: 0.4350 - val_accuracy: 0.8120

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0002.ckpt
Epoch 3/100
625/625 [==============================] - 80s 128ms/step - loss: 0.2239 - accuracy: 0.9120 - val_loss: 0.3559 - val_accuracy: 0.8455

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0003.ckpt
Epoch 4/100
625/625 [==============================] - 80s 128ms/step - loss: 0.1643 - accuracy: 0.9362 - val_loss: 0.2788 - val_accuracy: 0.8795

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0004.ckpt
Epoch 5/100
625/625 [==============================] - 79s 126ms/step - loss: 0.1315 - accuracy: 0.9489 - val_loss: 0.2229 - val_accuracy: 0.9107

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0005.ckpt
Epoch 6/100
625/625 [==============================] - 78s 125ms/step - loss: 0.1084 - accuracy: 0.9575 - val_loss: 0.2236 - val_accuracy: 0.9080

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0006.ckpt
Epoch 7/100
625/625 [==============================] - 79s 126ms/step - loss: 0.0961 - accuracy: 0.9638 - val_loss: 0.6200 - val_accuracy: 0.8652

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0007.ckpt
Epoch 8/100
625/625 [==============================] - 80s 128ms/step - loss: 0.1001 - accuracy: 0.9615 - val_loss: 0.1769 - val_accuracy: 0.9350

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0008.ckpt
Epoch 9/100
625/625 [==============================] - 79s 126ms/step - loss: 0.1278 - accuracy: 0.9484 - val_loss: 0.4366 - val_accuracy: 0.8183

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0009.ckpt
Epoch 10/100
625/625 [==============================] - 79s 127ms/step - loss: 0.2388 - accuracy: 0.9044 - val_loss: 0.1700 - val_accuracy: 0.9330

Epoch 00010: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0010.ckpt
Epoch 11/100
625/625 [==============================] - 82s 132ms/step - loss: 0.0936 - accuracy: 0.9654 - val_loss: 0.4674 - val_accuracy: 0.8263

Epoch 00011: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_1\cp-0011.ckpt

In [36]:

plot_model__hist(ResNet50V2_model_hist)
loss,acc = ResNet50V2_model.evaluate(test_set, verbose=2)
print("ResNet50V2_model(Adam) 백본 프리징 해제 후 모델의 정확도: {:5.2f}%".format(100*acc))
print("ResNet50V2_model(Adam) 백본 프리징 해제 후 모델의 Loss: {}".format(loss))

<Figure size 432x288 with 0 Axes>

100/100 - 6s - loss: 0.4755 - accuracy: 0.8240
ResNet50V2_model(Adam) 백본 프리징 해제 후 모델의 정확도: 82.40%
ResNet50V2_model(Adam) 백본 프리징 해제 후 모델의 Loss: 0.4754612147808075

Adam 모델백본 프리징 해제 후 학습시 95% -> 82.4%로 성능이 저하되었다.¶

• 수렴 전에 조기중단 된 것으로 판단.

경험에 따르면 이런 경우 SGD가 좋은 성능을 보였다 SGD로 학습해보자.¶

In [ ]:

train_set, val_set, test_set = get_dataset(size=128)
ResNet50V2_model_sgd = keras.Sequential([
    get_ResNet50V2(size=128,trainable=True), # 백본 학습 가능하게 설정
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])

checkpoint_path_3 = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path_3)


opt = SGD(lr=0.01, decay=1e-6, momentum=0.001,nesterov=True)
ResNet50V2_model_sgd.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

Found 20000 images belonging to 2 classes.
Found 4000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.

In [ ]:

ResNet50V2_model_sgd_hist = ResNet50V2_model_sgd.fit(train_set,
                    validation_data=val_set,
                    epochs=100,
                    callbacks = [callbacks])

Epoch 1/100
625/625 [==============================] - 84s 127ms/step - loss: 0.4003 - accuracy: 0.7940 - val_loss: 0.1246 - val_accuracy: 0.9503

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0001.ckpt
Epoch 2/100
625/625 [==============================] - 80s 127ms/step - loss: 0.1006 - accuracy: 0.9639 - val_loss: 0.1192 - val_accuracy: 0.9542

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0002.ckpt
Epoch 3/100
625/625 [==============================] - 80s 128ms/step - loss: 0.0464 - accuracy: 0.9840 - val_loss: 0.1191 - val_accuracy: 0.9563

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0003.ckpt
Epoch 4/100
625/625 [==============================] - 80s 128ms/step - loss: 0.0265 - accuracy: 0.9912 - val_loss: 0.1431 - val_accuracy: 0.9540

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0004.ckpt
Epoch 5/100
625/625 [==============================] - 80s 128ms/step - loss: 0.0185 - accuracy: 0.9946 - val_loss: 0.1432 - val_accuracy: 0.9592

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0005.ckpt
Epoch 6/100
625/625 [==============================] - 80s 128ms/step - loss: 0.0243 - accuracy: 0.9915 - val_loss: 0.1468 - val_accuracy: 0.9607

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0006.ckpt
Epoch 7/100
625/625 [==============================] - 81s 130ms/step - loss: 0.0138 - accuracy: 0.9956 - val_loss: 0.1537 - val_accuracy: 0.9588

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0007.ckpt
Epoch 8/100
625/625 [==============================] - 82s 131ms/step - loss: 0.0097 - accuracy: 0.9971 - val_loss: 0.1549 - val_accuracy: 0.9628

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0008.ckpt
Epoch 9/100
625/625 [==============================] - 82s 130ms/step - loss: 0.0092 - accuracy: 0.9974 - val_loss: 0.1848 - val_accuracy: 0.9578

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0009.ckpt
Epoch 10/100
625/625 [==============================] - 82s 130ms/step - loss: 0.0055 - accuracy: 0.9988 - val_loss: 0.1579 - val_accuracy: 0.9632

Epoch 00010: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0010.ckpt
Epoch 11/100
625/625 [==============================] - 81s 130ms/step - loss: 0.0030 - accuracy: 0.9995 - val_loss: 0.1776 - val_accuracy: 0.9615

Epoch 00011: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0011.ckpt
Epoch 12/100
625/625 [==============================] - 82s 130ms/step - loss: 0.0051 - accuracy: 0.9982 - val_loss: 0.1700 - val_accuracy: 0.9643

Epoch 00012: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0012.ckpt
Epoch 13/100
625/625 [==============================] - 82s 131ms/step - loss: 0.0036 - accuracy: 0.9985 - val_loss: 0.1924 - val_accuracy: 0.9620

Epoch 00013: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0013.ckpt
Epoch 14/100
625/625 [==============================] - 81s 130ms/step - loss: 0.0056 - accuracy: 0.9981 - val_loss: 0.1871 - val_accuracy: 0.9580

Epoch 00014: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0014.ckpt
Epoch 15/100
625/625 [==============================] - 82s 130ms/step - loss: 0.0039 - accuracy: 0.9989 - val_loss: 0.1852 - val_accuracy: 0.9628

Epoch 00015: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_2\cp-0015.ckpt

In [ ]:

plot_model__hist(ResNet50V2_model_sgd_hist)
loss_sgd,acc_sgd = ResNet50V2_model_sgd.evaluate(test_set, verbose=2)
print("ResNet50V2_model(SGD) 백본 프리징 해제 후 모델의 정확도: {:5.2f}%".format(100*acc_sgd))
print("ResNet50V2_model(SGD) 백본 프리징 해제 후 모델의 Loss: {}".format(loss_sgd))

<Figure size 432x288 with 0 Axes>

100/100 - 6s - loss: 0.1670 - accuracy: 0.9700
ResNet50V2_model(SGD) 백본 프리징 해제 후 모델의 정확도: 97.00%
ResNet50V2_model(SGD) 백본 프리징 해제 후 모델의 Loss: 0.16704358160495758

97%의 성능을 도출했다.¶

데이터 수 N은 Hoeffiding Inequality에 의해 모델의 성능과 비례함을 배웠다. 이미지 데이터 크롤링을 통해서 학습용 데이터를 추가확보한다.¶

• 크롤링 코드와 결과는 별도 python 파일과 ppt로 첨부

In [24]:

from Crawling_Naver import *

add_dataset_path = r"D:\Project(2)\cat&dog_set\additional_train"
crawl_dog = crawling_NAVER()
crawl_dog.crawling_start(keyword="개",index=2000, directory=add_dataset_path, category="dog_", forderName="dog")
crawl_cat = crawling_NAVER()
crawl_cat.crawling_start(keyword="고양이",index=2000, directory=add_dataset_path, category="cat_", forderName="cat")

Geometry
blur
Contrast
Color
ImgCorruptlike
Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[11]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[35]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[61]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[85]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[109]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[133]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[161]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[185]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[209]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[233]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[261]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[285]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[309]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[333]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[361]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[385]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[409]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Geometry
blur
Contrast
Color
ImgCorruptlike
Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[11]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[35]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[61]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[85]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[109]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[133]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[161]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[185]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[209]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[233]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[261]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[285]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[309]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Message: no such element: Unable to locate element: {"method":"xpath","selector":"//*[@id="main_pack"]/section/div/div[1]/div[1]/div[333]/div/div[1]/a/img"}
  (Session info: chrome=93.0.4577.63)

Out[24]:

True

크롤링한 데이터로 이미지 증강 거쳐서 데이터셋 재구성한 후 SGD 학습¶

• 97.5%로 0.5%의 성능향상

In [9]:

train_set, val_set, test_set = get_dataset(crawl=True,aug=True,size=128)
# SGD 모델 
ResNet50V2_model_sgd_crawl = keras.Sequential([
    get_ResNet50V2(size=128,trainable=True), # 백본 학습 가능하게 설정
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])
# 가중치 저장경로 변경
checkpoint_path = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path,patience=5)#동일 조건


opt = SGD(lr=0.01, decay=1e-6, momentum=0.001,nesterov=True)
ResNet50V2_model_sgd_crawl.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

ResNet50V2_model_sgd_crawl_hist = ResNet50V2_model_sgd_crawl.fit(train_set,
                    validation_data=val_set,
                    epochs=100,
                    callbacks = [callbacks])

Found 21900 images belonging to 2 classes.
Found 4000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.
Epoch 1/100
685/685 [==============================] - 237s 337ms/step - loss: 0.4919 - accuracy: 0.7392 - val_loss: 0.1578 - val_accuracy: 0.9362

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0001.ckpt
Epoch 2/100
685/685 [==============================] - 111s 162ms/step - loss: 0.2244 - accuracy: 0.9087 - val_loss: 0.1378 - val_accuracy: 0.9445

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0002.ckpt
Epoch 3/100
685/685 [==============================] - 113s 165ms/step - loss: 0.1749 - accuracy: 0.9329 - val_loss: 0.1892 - val_accuracy: 0.9230

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0003.ckpt
Epoch 4/100
685/685 [==============================] - 111s 162ms/step - loss: 0.1392 - accuracy: 0.9486 - val_loss: 0.1417 - val_accuracy: 0.9470

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0004.ckpt
Epoch 5/100
685/685 [==============================] - 108s 157ms/step - loss: 0.1208 - accuracy: 0.9533 - val_loss: 0.1082 - val_accuracy: 0.9567

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0005.ckpt
Epoch 6/100
685/685 [==============================] - 112s 163ms/step - loss: 0.1131 - accuracy: 0.9594 - val_loss: 0.1126 - val_accuracy: 0.9532

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0006.ckpt
Epoch 7/100
685/685 [==============================] - 115s 167ms/step - loss: 0.0921 - accuracy: 0.9663 - val_loss: 0.1147 - val_accuracy: 0.9597

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0007.ckpt
Epoch 8/100
685/685 [==============================] - 114s 166ms/step - loss: 0.0859 - accuracy: 0.9671 - val_loss: 0.1083 - val_accuracy: 0.9610

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0008.ckpt
Epoch 9/100
685/685 [==============================] - 118s 172ms/step - loss: 0.0857 - accuracy: 0.9688 - val_loss: 0.1157 - val_accuracy: 0.9567

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0009.ckpt
Epoch 10/100
685/685 [==============================] - 110s 160ms/step - loss: 0.0714 - accuracy: 0.9730 - val_loss: 0.1000 - val_accuracy: 0.9625

Epoch 00010: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0010.ckpt
Epoch 11/100
685/685 [==============================] - 109s 159ms/step - loss: 0.0706 - accuracy: 0.9753 - val_loss: 0.1534 - val_accuracy: 0.9480

Epoch 00011: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0011.ckpt
Epoch 12/100
685/685 [==============================] - 108s 157ms/step - loss: 0.0588 - accuracy: 0.9796 - val_loss: 0.1014 - val_accuracy: 0.9628

Epoch 00012: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0012.ckpt
Epoch 13/100
685/685 [==============================] - 112s 163ms/step - loss: 0.0525 - accuracy: 0.9813 - val_loss: 0.1365 - val_accuracy: 0.9575

Epoch 00013: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0013.ckpt
Epoch 14/100
685/685 [==============================] - 110s 160ms/step - loss: 0.0539 - accuracy: 0.9807 - val_loss: 0.1091 - val_accuracy: 0.9632

Epoch 00014: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0014.ckpt
Epoch 15/100
685/685 [==============================] - 111s 162ms/step - loss: 0.0540 - accuracy: 0.9806 - val_loss: 0.1008 - val_accuracy: 0.9660

Epoch 00015: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0015.ckpt
Epoch 16/100
685/685 [==============================] - 111s 162ms/step - loss: 0.0470 - accuracy: 0.9830 - val_loss: 0.1152 - val_accuracy: 0.9595

Epoch 00016: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0016.ckpt
Epoch 17/100
685/685 [==============================] - 107s 157ms/step - loss: 0.0460 - accuracy: 0.9841 - val_loss: 0.1203 - val_accuracy: 0.9613

Epoch 00017: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0017.ckpt
Epoch 18/100
685/685 [==============================] - 110s 160ms/step - loss: 0.0390 - accuracy: 0.9855 - val_loss: 0.1396 - val_accuracy: 0.9567

Epoch 00018: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0018.ckpt
Epoch 19/100
685/685 [==============================] - 108s 158ms/step - loss: 0.0385 - accuracy: 0.9871 - val_loss: 0.1227 - val_accuracy: 0.9610

Epoch 00019: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0019.ckpt
Epoch 20/100
685/685 [==============================] - 110s 160ms/step - loss: 0.0374 - accuracy: 0.9870 - val_loss: 0.1113 - val_accuracy: 0.9660

Epoch 00020: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0020.ckpt

In [10]:

plot_model__hist(ResNet50V2_model_sgd_crawl_hist)
loss_sgd,acc_sgd = ResNet50V2_model_sgd_crawl.evaluate(test_set, verbose=2)
print("ResNet50V2_model(SGD) 이미지 크롤링,증강 후 모델의 정확도: {:5.2f}%".format(100*acc_sgd))
print("ResNet50V2_model(SGD) 이미지 크롤링,증강 후 모델의 Loss: {}".format(loss_sgd))

<Figure size 432x288 with 0 Axes>

100/100 - 8s - loss: 0.0836 - accuracy: 0.9750
ResNet50V2_model(SGD) 이미지 크롤링,증강 후 모델의 정확도: 97.50%
ResNet50V2_model(SGD) 이미지 크롤링,증강 후 모델의 Loss: 0.083615742623806

성능이 좋지 않았던 Adam으로 early stopping patience 50으로 수렴에 가까워질 때 까지 학습하기¶

• 96.1% 의 성능 도출

In [23]:

from tensorflow.keras.optimizers import Adam
train_set, val_set, test_set = get_dataset(crawl=True,aug=True,size=128)
# Adam 모델 
ResNet50V2_model_adam_crawl = keras.Sequential([
    get_ResNet50V2(size=128,trainable=True), # 백본 학습 가능하게 설정
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])
# 가중치 저장경로 변경
checkpoint_path = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path,patience=50)#동일 조건


opt = Adam(learning_rate=0.001)
ResNet50V2_model_adam_crawl.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

ResNet50V2_model_adam_crawl_hist = ResNet50V2_model_adam_crawl.fit(train_set,
                    validation_data=val_set,
                    epochs=500,
                    callbacks = [callbacks])

Found 20000 images belonging to 2 classes.
Found 4000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.
Epoch 1/500
625/625 [==============================] - 202s 317ms/step - loss: 0.6546 - accuracy: 0.6484 - val_loss: 0.6594 - val_accuracy: 0.7147

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0001.ckpt
Epoch 2/500
625/625 [==============================] - 104s 166ms/step - loss: 0.5193 - accuracy: 0.7574 - val_loss: 0.6010 - val_accuracy: 0.7243

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0002.ckpt
Epoch 3/500
625/625 [==============================] - 104s 166ms/step - loss: 0.4368 - accuracy: 0.8022 - val_loss: 0.7599 - val_accuracy: 0.6888

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0003.ckpt
Epoch 4/500
625/625 [==============================] - 104s 166ms/step - loss: 0.3874 - accuracy: 0.8318 - val_loss: 0.5589 - val_accuracy: 0.8027

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0004.ckpt
Epoch 5/500
625/625 [==============================] - 104s 166ms/step - loss: 0.3638 - accuracy: 0.8490 - val_loss: 0.6252 - val_accuracy: 0.7490

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0005.ckpt
Epoch 6/500
625/625 [==============================] - 105s 168ms/step - loss: 0.3249 - accuracy: 0.8627 - val_loss: 0.4346 - val_accuracy: 0.7883

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0006.ckpt
Epoch 7/500
625/625 [==============================] - 104s 166ms/step - loss: 0.3152 - accuracy: 0.8654 - val_loss: 0.3762 - val_accuracy: 0.8455

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0007.ckpt
Epoch 8/500
625/625 [==============================] - 104s 166ms/step - loss: 0.2858 - accuracy: 0.8799 - val_loss: 0.4503 - val_accuracy: 0.8120

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0008.ckpt
Epoch 9/500
625/625 [==============================] - 104s 166ms/step - loss: 0.3160 - accuracy: 0.8656 - val_loss: 0.2513 - val_accuracy: 0.8985

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0009.ckpt
Epoch 10/500
625/625 [==============================] - 104s 167ms/step - loss: 0.2629 - accuracy: 0.8916 - val_loss: 0.3425 - val_accuracy: 0.8370

Epoch 00010: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0010.ckpt
Epoch 11/500
625/625 [==============================] - 104s 166ms/step - loss: 0.3055 - accuracy: 0.8748 - val_loss: 0.3927 - val_accuracy: 0.8307

Epoch 00011: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0011.ckpt
Epoch 12/500
625/625 [==============================] - 106s 170ms/step - loss: 0.2610 - accuracy: 0.8908 - val_loss: 0.8189 - val_accuracy: 0.6535

Epoch 00012: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0012.ckpt
Epoch 13/500
625/625 [==============================] - 104s 166ms/step - loss: 0.2555 - accuracy: 0.8949 - val_loss: 0.2603 - val_accuracy: 0.8865

Epoch 00013: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0013.ckpt
Epoch 14/500
625/625 [==============================] - 104s 166ms/step - loss: 0.2233 - accuracy: 0.9058 - val_loss: 0.2366 - val_accuracy: 0.9047

Epoch 00014: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0014.ckpt
Epoch 15/500
625/625 [==============================] - 105s 168ms/step - loss: 0.2176 - accuracy: 0.9143 - val_loss: 0.3208 - val_accuracy: 0.8447

Epoch 00015: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0015.ckpt
Epoch 16/500
625/625 [==============================] - 105s 168ms/step - loss: 0.2697 - accuracy: 0.8875 - val_loss: 0.4351 - val_accuracy: 0.8428

Epoch 00016: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0016.ckpt
Epoch 17/500
625/625 [==============================] - 108s 172ms/step - loss: 0.2640 - accuracy: 0.8890 - val_loss: 0.2558 - val_accuracy: 0.8913

Epoch 00017: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0017.ckpt
Epoch 18/500
625/625 [==============================] - 105s 168ms/step - loss: 0.2393 - accuracy: 0.9059 - val_loss: 0.2567 - val_accuracy: 0.8810

Epoch 00018: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0018.ckpt
Epoch 19/500
625/625 [==============================] - 106s 169ms/step - loss: 0.2181 - accuracy: 0.9114 - val_loss: 0.3141 - val_accuracy: 0.8593

Epoch 00019: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0019.ckpt
Epoch 20/500
625/625 [==============================] - 105s 168ms/step - loss: 0.2157 - accuracy: 0.9196 - val_loss: 0.1916 - val_accuracy: 0.9222

Epoch 00020: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0020.ckpt
Epoch 21/500
625/625 [==============================] - 106s 169ms/step - loss: 0.1864 - accuracy: 0.9270 - val_loss: 0.2959 - val_accuracy: 0.8712

Epoch 00021: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0021.ckpt
Epoch 22/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1806 - accuracy: 0.9287 - val_loss: 0.2371 - val_accuracy: 0.9078

Epoch 00022: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0022.ckpt
Epoch 23/500
625/625 [==============================] - 107s 172ms/step - loss: 0.1842 - accuracy: 0.9273 - val_loss: 0.2146 - val_accuracy: 0.9120

Epoch 00023: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0023.ckpt
Epoch 24/500
625/625 [==============================] - 107s 171ms/step - loss: 0.1710 - accuracy: 0.9325 - val_loss: 0.3586 - val_accuracy: 0.8715

Epoch 00024: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0024.ckpt
Epoch 25/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1955 - accuracy: 0.9224 - val_loss: 0.2243 - val_accuracy: 0.9087

Epoch 00025: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0025.ckpt
Epoch 26/500
625/625 [==============================] - 106s 169ms/step - loss: 0.1736 - accuracy: 0.9307 - val_loss: 0.4214 - val_accuracy: 0.8313

Epoch 00026: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0026.ckpt
Epoch 27/500
625/625 [==============================] - 105s 169ms/step - loss: 0.2260 - accuracy: 0.9108 - val_loss: 0.2034 - val_accuracy: 0.9195

Epoch 00027: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0027.ckpt
Epoch 28/500
625/625 [==============================] - 106s 169ms/step - loss: 0.1654 - accuracy: 0.9331 - val_loss: 1.0227 - val_accuracy: 0.5900

Epoch 00028: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0028.ckpt
Epoch 29/500
625/625 [==============================] - 106s 170ms/step - loss: 0.1936 - accuracy: 0.9235 - val_loss: 0.1998 - val_accuracy: 0.9150

Epoch 00029: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0029.ckpt
Epoch 30/500
625/625 [==============================] - 106s 169ms/step - loss: 0.1540 - accuracy: 0.9390 - val_loss: 0.5147 - val_accuracy: 0.7600

Epoch 00030: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0030.ckpt
Epoch 31/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1644 - accuracy: 0.9352 - val_loss: 0.2664 - val_accuracy: 0.9038

Epoch 00031: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0031.ckpt
Epoch 32/500
625/625 [==============================] - 106s 169ms/step - loss: 0.1491 - accuracy: 0.9393 - val_loss: 0.2045 - val_accuracy: 0.9212

Epoch 00032: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0032.ckpt
Epoch 33/500
625/625 [==============================] - 106s 169ms/step - loss: 0.1487 - accuracy: 0.9411 - val_loss: 0.2565 - val_accuracy: 0.9078

Epoch 00033: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0033.ckpt
Epoch 34/500
625/625 [==============================] - 108s 172ms/step - loss: 0.1579 - accuracy: 0.9376 - val_loss: 0.2028 - val_accuracy: 0.9233

Epoch 00034: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0034.ckpt
Epoch 35/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1388 - accuracy: 0.9440 - val_loss: 0.2212 - val_accuracy: 0.9015

Epoch 00035: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0035.ckpt
Epoch 36/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1583 - accuracy: 0.9370 - val_loss: 0.1724 - val_accuracy: 0.9270

Epoch 00036: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0036.ckpt
Epoch 37/500
625/625 [==============================] - 106s 169ms/step - loss: 0.1296 - accuracy: 0.9481 - val_loss: 0.1730 - val_accuracy: 0.9358

Epoch 00037: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0037.ckpt
Epoch 38/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1322 - accuracy: 0.9500 - val_loss: 0.1665 - val_accuracy: 0.9350

Epoch 00038: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0038.ckpt
Epoch 39/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1233 - accuracy: 0.9519 - val_loss: 0.2274 - val_accuracy: 0.9137

Epoch 00039: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0039.ckpt
Epoch 40/500
625/625 [==============================] - 107s 171ms/step - loss: 0.1224 - accuracy: 0.9527 - val_loss: 0.1730 - val_accuracy: 0.9315

Epoch 00040: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0040.ckpt
Epoch 41/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1221 - accuracy: 0.9538 - val_loss: 0.2472 - val_accuracy: 0.8995

Epoch 00041: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0041.ckpt
Epoch 42/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1234 - accuracy: 0.9525 - val_loss: 0.1568 - val_accuracy: 0.9373

Epoch 00042: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0042.ckpt
Epoch 43/500
625/625 [==============================] - 105s 169ms/step - loss: 0.1123 - accuracy: 0.9548 - val_loss: 0.1778 - val_accuracy: 0.9237

Epoch 00043: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0043.ckpt
Epoch 44/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1122 - accuracy: 0.9573 - val_loss: 0.2323 - val_accuracy: 0.9162

Epoch 00044: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0044.ckpt
Epoch 45/500
625/625 [==============================] - 107s 170ms/step - loss: 0.1227 - accuracy: 0.9509 - val_loss: 0.2668 - val_accuracy: 0.8863

Epoch 00045: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0045.ckpt
Epoch 46/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1157 - accuracy: 0.9554 - val_loss: 0.1586 - val_accuracy: 0.9345

Epoch 00046: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0046.ckpt
Epoch 47/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1002 - accuracy: 0.9621 - val_loss: 0.2078 - val_accuracy: 0.9237

Epoch 00047: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0047.ckpt
Epoch 48/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1066 - accuracy: 0.9592 - val_loss: 0.4060 - val_accuracy: 0.8175

Epoch 00048: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0048.ckpt
Epoch 49/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1049 - accuracy: 0.9585 - val_loss: 0.2185 - val_accuracy: 0.9233

Epoch 00049: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0049.ckpt
Epoch 50/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0966 - accuracy: 0.9610 - val_loss: 0.3210 - val_accuracy: 0.8630

Epoch 00050: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0050.ckpt
Epoch 51/500
625/625 [==============================] - 107s 172ms/step - loss: 0.1017 - accuracy: 0.9598 - val_loss: 0.2293 - val_accuracy: 0.9218

Epoch 00051: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0051.ckpt
Epoch 52/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0920 - accuracy: 0.9664 - val_loss: 0.5627 - val_accuracy: 0.8235

Epoch 00052: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0052.ckpt
Epoch 53/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1009 - accuracy: 0.9612 - val_loss: 0.2185 - val_accuracy: 0.9120

Epoch 00053: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0053.ckpt
Epoch 54/500
625/625 [==============================] - 106s 169ms/step - loss: 0.0919 - accuracy: 0.9656 - val_loss: 0.2101 - val_accuracy: 0.9150

Epoch 00054: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0054.ckpt
Epoch 55/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0936 - accuracy: 0.9620 - val_loss: 0.1538 - val_accuracy: 0.9433

Epoch 00055: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0055.ckpt
Epoch 56/500
625/625 [==============================] - 106s 169ms/step - loss: 0.0837 - accuracy: 0.9686 - val_loss: 0.2299 - val_accuracy: 0.9370

Epoch 00056: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0056.ckpt
Epoch 57/500
625/625 [==============================] - 111s 177ms/step - loss: 0.1115 - accuracy: 0.9546 - val_loss: 0.2215 - val_accuracy: 0.9130

Epoch 00057: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0057.ckpt
Epoch 58/500
625/625 [==============================] - 105s 167ms/step - loss: 0.1702 - accuracy: 0.9345 - val_loss: 0.1531 - val_accuracy: 0.9385

Epoch 00058: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0058.ckpt
Epoch 59/500
625/625 [==============================] - 105s 167ms/step - loss: 0.1170 - accuracy: 0.9550 - val_loss: 0.1769 - val_accuracy: 0.9358

Epoch 00059: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0059.ckpt
Epoch 60/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0940 - accuracy: 0.9628 - val_loss: 0.1408 - val_accuracy: 0.9442

Epoch 00060: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0060.ckpt
Epoch 61/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0820 - accuracy: 0.9688 - val_loss: 0.1525 - val_accuracy: 0.9463

Epoch 00061: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0061.ckpt
Epoch 62/500
625/625 [==============================] - 107s 170ms/step - loss: 0.0799 - accuracy: 0.9696 - val_loss: 0.1426 - val_accuracy: 0.9457

Epoch 00062: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0062.ckpt
Epoch 63/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0756 - accuracy: 0.9713 - val_loss: 0.1500 - val_accuracy: 0.9435

Epoch 00063: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0063.ckpt
Epoch 64/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0834 - accuracy: 0.9693 - val_loss: 0.2452 - val_accuracy: 0.9165

Epoch 00064: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0064.ckpt
Epoch 65/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0781 - accuracy: 0.9716 - val_loss: 0.1516 - val_accuracy: 0.9405

Epoch 00065: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0065.ckpt
Epoch 66/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0738 - accuracy: 0.9730 - val_loss: 0.3240 - val_accuracy: 0.8783

Epoch 00066: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0066.ckpt
Epoch 67/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0767 - accuracy: 0.9713 - val_loss: 0.2161 - val_accuracy: 0.9193

Epoch 00067: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0067.ckpt
Epoch 68/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0784 - accuracy: 0.9706 - val_loss: 0.2817 - val_accuracy: 0.9000

Epoch 00068: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0068.ckpt
Epoch 69/500
625/625 [==============================] - 104s 166ms/step - loss: 0.0807 - accuracy: 0.9688 - val_loss: 0.1969 - val_accuracy: 0.9268

Epoch 00069: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0069.ckpt
Epoch 70/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0732 - accuracy: 0.9723 - val_loss: 0.2965 - val_accuracy: 0.8925

Epoch 00070: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0070.ckpt
Epoch 71/500
625/625 [==============================] - 104s 166ms/step - loss: 0.0717 - accuracy: 0.9733 - val_loss: 0.2526 - val_accuracy: 0.9118

Epoch 00071: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0071.ckpt
Epoch 72/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0679 - accuracy: 0.9748 - val_loss: 0.2147 - val_accuracy: 0.9210

Epoch 00072: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0072.ckpt
Epoch 73/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0676 - accuracy: 0.9748 - val_loss: 0.1885 - val_accuracy: 0.9330

Epoch 00073: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0073.ckpt
Epoch 74/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0708 - accuracy: 0.9720 - val_loss: 0.1928 - val_accuracy: 0.9325

Epoch 00074: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0074.ckpt
Epoch 75/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0638 - accuracy: 0.9753 - val_loss: 0.3218 - val_accuracy: 0.8995

Epoch 00075: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0075.ckpt
Epoch 76/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0668 - accuracy: 0.9756 - val_loss: 0.2725 - val_accuracy: 0.9060

Epoch 00076: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0076.ckpt
Epoch 77/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0680 - accuracy: 0.9731 - val_loss: 0.1870 - val_accuracy: 0.9380

Epoch 00077: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0077.ckpt
Epoch 78/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0668 - accuracy: 0.9754 - val_loss: 0.2758 - val_accuracy: 0.9262

Epoch 00078: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0078.ckpt
Epoch 79/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0645 - accuracy: 0.9760 - val_loss: 0.1899 - val_accuracy: 0.9435

Epoch 00079: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0079.ckpt
Epoch 80/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0599 - accuracy: 0.9768 - val_loss: 0.2627 - val_accuracy: 0.9185

Epoch 00080: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0080.ckpt
Epoch 81/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0595 - accuracy: 0.9784 - val_loss: 0.2172 - val_accuracy: 0.9367

Epoch 00081: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0081.ckpt
Epoch 82/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0610 - accuracy: 0.9780 - val_loss: 0.2161 - val_accuracy: 0.9335

Epoch 00082: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0082.ckpt
Epoch 83/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0615 - accuracy: 0.9737 - val_loss: 0.1572 - val_accuracy: 0.9435

Epoch 00083: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0083.ckpt
Epoch 84/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0578 - accuracy: 0.9763 - val_loss: 0.1662 - val_accuracy: 0.9490

Epoch 00084: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0084.ckpt
Epoch 85/500
625/625 [==============================] - 107s 171ms/step - loss: 0.0548 - accuracy: 0.9779 - val_loss: 0.2449 - val_accuracy: 0.9087

Epoch 00085: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0085.ckpt
Epoch 86/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0527 - accuracy: 0.9804 - val_loss: 0.2897 - val_accuracy: 0.8873

Epoch 00086: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0086.ckpt
Epoch 87/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0549 - accuracy: 0.9782 - val_loss: 0.1784 - val_accuracy: 0.9488

Epoch 00087: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0087.ckpt
Epoch 88/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0495 - accuracy: 0.9819 - val_loss: 0.1811 - val_accuracy: 0.9392

Epoch 00088: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0088.ckpt
Epoch 89/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0586 - accuracy: 0.9797 - val_loss: 0.2152 - val_accuracy: 0.9350

Epoch 00089: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0089.ckpt
Epoch 90/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0523 - accuracy: 0.9813 - val_loss: 0.4403 - val_accuracy: 0.8863

Epoch 00090: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0090.ckpt
Epoch 91/500
625/625 [==============================] - 109s 175ms/step - loss: 0.0483 - accuracy: 0.9828 - val_loss: 0.2882 - val_accuracy: 0.9187

Epoch 00091: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0091.ckpt
Epoch 92/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0562 - accuracy: 0.9779 - val_loss: 0.2222 - val_accuracy: 0.9345

Epoch 00092: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0092.ckpt
Epoch 93/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0513 - accuracy: 0.9797 - val_loss: 0.1993 - val_accuracy: 0.9310

Epoch 00093: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0093.ckpt
Epoch 94/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0549 - accuracy: 0.9796 - val_loss: 0.1915 - val_accuracy: 0.9477

Epoch 00094: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0094.ckpt
Epoch 95/500
625/625 [==============================] - 113s 180ms/step - loss: 0.0576 - accuracy: 0.9805 - val_loss: 0.2244 - val_accuracy: 0.9463

Epoch 00095: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0095.ckpt
Epoch 96/500
625/625 [==============================] - 114s 183ms/step - loss: 0.0550 - accuracy: 0.9800 - val_loss: 0.3378 - val_accuracy: 0.8985

Epoch 00096: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0096.ckpt
Epoch 97/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0441 - accuracy: 0.9840 - val_loss: 0.1558 - val_accuracy: 0.9473

Epoch 00097: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0097.ckpt
Epoch 98/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0437 - accuracy: 0.9847 - val_loss: 0.1626 - val_accuracy: 0.9448

Epoch 00098: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0098.ckpt
Epoch 99/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0464 - accuracy: 0.9832 - val_loss: 0.2103 - val_accuracy: 0.9333

Epoch 00099: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0099.ckpt
Epoch 100/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0484 - accuracy: 0.9819 - val_loss: 0.1700 - val_accuracy: 0.9475

Epoch 00100: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0100.ckpt
Epoch 101/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0444 - accuracy: 0.9833 - val_loss: 0.2528 - val_accuracy: 0.9268

Epoch 00101: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0101.ckpt
Epoch 102/500
625/625 [==============================] - 107s 171ms/step - loss: 0.0483 - accuracy: 0.9823 - val_loss: 0.2666 - val_accuracy: 0.9233

Epoch 00102: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0102.ckpt
Epoch 103/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0413 - accuracy: 0.9842 - val_loss: 0.1725 - val_accuracy: 0.9490

Epoch 00103: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0103.ckpt
Epoch 104/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0437 - accuracy: 0.9829 - val_loss: 0.2275 - val_accuracy: 0.9237

Epoch 00104: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0104.ckpt
Epoch 105/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0682 - accuracy: 0.9772 - val_loss: 0.1648 - val_accuracy: 0.9365

Epoch 00105: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0105.ckpt
Epoch 106/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0876 - accuracy: 0.9676 - val_loss: 0.1563 - val_accuracy: 0.9500

Epoch 00106: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0106.ckpt
Epoch 107/500
625/625 [==============================] - 106s 169ms/step - loss: 0.0443 - accuracy: 0.9831 - val_loss: 0.1679 - val_accuracy: 0.9535

Epoch 00107: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0107.ckpt
Epoch 108/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0406 - accuracy: 0.9861 - val_loss: 0.1549 - val_accuracy: 0.9507

Epoch 00108: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0108.ckpt
Epoch 109/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0333 - accuracy: 0.9883 - val_loss: 0.2104 - val_accuracy: 0.9390

Epoch 00109: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0109.ckpt
Epoch 110/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0313 - accuracy: 0.9880 - val_loss: 0.2171 - val_accuracy: 0.9352

Epoch 00110: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0110.ckpt
Epoch 111/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0322 - accuracy: 0.9878 - val_loss: 0.2522 - val_accuracy: 0.9438

Epoch 00111: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0111.ckpt
Epoch 112/500
625/625 [==============================] - 104s 166ms/step - loss: 0.0377 - accuracy: 0.9870 - val_loss: 0.2101 - val_accuracy: 0.9400

Epoch 00112: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0112.ckpt
Epoch 113/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0369 - accuracy: 0.9847 - val_loss: 0.2389 - val_accuracy: 0.9312

Epoch 00113: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0113.ckpt
Epoch 114/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0402 - accuracy: 0.9853 - val_loss: 0.2874 - val_accuracy: 0.9028

Epoch 00114: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0114.ckpt
Epoch 115/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0387 - accuracy: 0.9862 - val_loss: 0.5950 - val_accuracy: 0.8512

Epoch 00115: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0115.ckpt
Epoch 116/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0402 - accuracy: 0.9858 - val_loss: 0.1802 - val_accuracy: 0.9498

Epoch 00116: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0116.ckpt
Epoch 117/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0498 - accuracy: 0.9814 - val_loss: 0.2324 - val_accuracy: 0.9330

Epoch 00117: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0117.ckpt
Epoch 118/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0384 - accuracy: 0.9862 - val_loss: 0.4130 - val_accuracy: 0.8850

Epoch 00118: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0118.ckpt
Epoch 119/500
625/625 [==============================] - 108s 172ms/step - loss: 0.0365 - accuracy: 0.9870 - val_loss: 0.1825 - val_accuracy: 0.9440

Epoch 00119: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0119.ckpt
Epoch 120/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0365 - accuracy: 0.9856 - val_loss: 0.1991 - val_accuracy: 0.9450

Epoch 00120: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0120.ckpt
Epoch 121/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0374 - accuracy: 0.9870 - val_loss: 0.2471 - val_accuracy: 0.9180

Epoch 00121: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0121.ckpt
Epoch 122/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0439 - accuracy: 0.9827 - val_loss: 0.2234 - val_accuracy: 0.9427

Epoch 00122: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0122.ckpt
Epoch 123/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0373 - accuracy: 0.9869 - val_loss: 0.2133 - val_accuracy: 0.9417

Epoch 00123: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0123.ckpt
Epoch 124/500
625/625 [==============================] - 105s 169ms/step - loss: 0.0373 - accuracy: 0.9867 - val_loss: 0.1884 - val_accuracy: 0.9467

Epoch 00124: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0124.ckpt
Epoch 125/500
625/625 [==============================] - 108s 172ms/step - loss: 0.0444 - accuracy: 0.9858 - val_loss: 0.2377 - val_accuracy: 0.9290

Epoch 00125: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0125.ckpt
Epoch 126/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0317 - accuracy: 0.9883 - val_loss: 0.2489 - val_accuracy: 0.9258

Epoch 00126: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0126.ckpt
Epoch 127/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0337 - accuracy: 0.9893 - val_loss: 0.2841 - val_accuracy: 0.9230

Epoch 00127: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0127.ckpt
Epoch 128/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0359 - accuracy: 0.9858 - val_loss: 0.1829 - val_accuracy: 0.9490

Epoch 00128: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0128.ckpt
Epoch 129/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0351 - accuracy: 0.9881 - val_loss: 0.1904 - val_accuracy: 0.9513

Epoch 00129: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0129.ckpt
Epoch 130/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0362 - accuracy: 0.9885 - val_loss: 0.2333 - val_accuracy: 0.9310

Epoch 00130: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0130.ckpt
Epoch 131/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0340 - accuracy: 0.9877 - val_loss: 0.1658 - val_accuracy: 0.9528

Epoch 00131: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0131.ckpt
Epoch 132/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0312 - accuracy: 0.9879 - val_loss: 0.1793 - val_accuracy: 0.9438

Epoch 00132: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0132.ckpt
Epoch 133/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0337 - accuracy: 0.9874 - val_loss: 0.1878 - val_accuracy: 0.9425

Epoch 00133: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0133.ckpt
Epoch 134/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0299 - accuracy: 0.9889 - val_loss: 0.2172 - val_accuracy: 0.9417

Epoch 00134: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0134.ckpt
Epoch 135/500
625/625 [==============================] - 105s 168ms/step - loss: 0.1117 - accuracy: 0.9659 - val_loss: 0.1820 - val_accuracy: 0.9302

Epoch 00135: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0135.ckpt
Epoch 136/500
625/625 [==============================] - 106s 170ms/step - loss: 0.1009 - accuracy: 0.9596 - val_loss: 0.1620 - val_accuracy: 0.9467

Epoch 00136: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0136.ckpt
Epoch 137/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0517 - accuracy: 0.9825 - val_loss: 0.1605 - val_accuracy: 0.9510

Epoch 00137: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0137.ckpt
Epoch 138/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0361 - accuracy: 0.9871 - val_loss: 0.1805 - val_accuracy: 0.9595

Epoch 00138: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0138.ckpt
Epoch 139/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0299 - accuracy: 0.9898 - val_loss: 0.2388 - val_accuracy: 0.9500

Epoch 00139: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0139.ckpt
Epoch 140/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0281 - accuracy: 0.9905 - val_loss: 0.2196 - val_accuracy: 0.9465

Epoch 00140: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0140.ckpt
Epoch 141/500
625/625 [==============================] - 106s 169ms/step - loss: 0.0262 - accuracy: 0.9906 - val_loss: 0.2247 - val_accuracy: 0.9455

Epoch 00141: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0141.ckpt
Epoch 142/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0194 - accuracy: 0.9941 - val_loss: 0.2295 - val_accuracy: 0.9485

Epoch 00142: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0142.ckpt
Epoch 143/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0220 - accuracy: 0.9926 - val_loss: 0.2068 - val_accuracy: 0.9445

Epoch 00143: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0143.ckpt
Epoch 144/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0244 - accuracy: 0.9909 - val_loss: 0.1596 - val_accuracy: 0.9560

Epoch 00144: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0144.ckpt
Epoch 145/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0262 - accuracy: 0.9913 - val_loss: 0.1860 - val_accuracy: 0.9467

Epoch 00145: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0145.ckpt
Epoch 146/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0271 - accuracy: 0.9883 - val_loss: 0.3543 - val_accuracy: 0.9290

Epoch 00146: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0146.ckpt
Epoch 147/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0324 - accuracy: 0.9879 - val_loss: 0.3320 - val_accuracy: 0.9258

Epoch 00147: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0147.ckpt
Epoch 148/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0302 - accuracy: 0.9893 - val_loss: 0.3184 - val_accuracy: 0.9280

Epoch 00148: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0148.ckpt
Epoch 149/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0338 - accuracy: 0.9870 - val_loss: 0.2088 - val_accuracy: 0.9442

Epoch 00149: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0149.ckpt
Epoch 150/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0244 - accuracy: 0.9916 - val_loss: 0.2526 - val_accuracy: 0.9345

Epoch 00150: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0150.ckpt
Epoch 151/500
625/625 [==============================] - 104s 166ms/step - loss: 0.0274 - accuracy: 0.9898 - val_loss: 0.2173 - val_accuracy: 0.9475

Epoch 00151: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0151.ckpt
Epoch 152/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0278 - accuracy: 0.9908 - val_loss: 0.2885 - val_accuracy: 0.9137

Epoch 00152: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0152.ckpt
Epoch 153/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0315 - accuracy: 0.9892 - val_loss: 0.3286 - val_accuracy: 0.9128

Epoch 00153: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0153.ckpt
Epoch 154/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0334 - accuracy: 0.9893 - val_loss: 0.2239 - val_accuracy: 0.9425

Epoch 00154: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0154.ckpt
Epoch 155/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0294 - accuracy: 0.9897 - val_loss: 0.5084 - val_accuracy: 0.8650

Epoch 00155: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0155.ckpt
Epoch 156/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0295 - accuracy: 0.9899 - val_loss: 0.1657 - val_accuracy: 0.9505

Epoch 00156: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0156.ckpt
Epoch 157/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0335 - accuracy: 0.9888 - val_loss: 0.2306 - val_accuracy: 0.9417

Epoch 00157: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0157.ckpt
Epoch 158/500
625/625 [==============================] - 107s 171ms/step - loss: 0.0279 - accuracy: 0.9909 - val_loss: 0.2741 - val_accuracy: 0.9287

Epoch 00158: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0158.ckpt
Epoch 159/500
625/625 [==============================] - 108s 172ms/step - loss: 0.0296 - accuracy: 0.9896 - val_loss: 0.2595 - val_accuracy: 0.9352

Epoch 00159: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0159.ckpt
Epoch 160/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0247 - accuracy: 0.9915 - val_loss: 0.2616 - val_accuracy: 0.9405

Epoch 00160: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0160.ckpt
Epoch 161/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0260 - accuracy: 0.9927 - val_loss: 0.2511 - val_accuracy: 0.9283

Epoch 00161: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0161.ckpt
Epoch 162/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0268 - accuracy: 0.9902 - val_loss: 0.2517 - val_accuracy: 0.9427

Epoch 00162: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0162.ckpt
Epoch 163/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0316 - accuracy: 0.9891 - val_loss: 0.1930 - val_accuracy: 0.9457

Epoch 00163: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0163.ckpt
Epoch 164/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0258 - accuracy: 0.9900 - val_loss: 0.2445 - val_accuracy: 0.9440

Epoch 00164: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0164.ckpt
Epoch 165/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0284 - accuracy: 0.9911 - val_loss: 0.2076 - val_accuracy: 0.9450

Epoch 00165: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0165.ckpt
Epoch 166/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0291 - accuracy: 0.9890 - val_loss: 0.2824 - val_accuracy: 0.9262

Epoch 00166: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0166.ckpt
Epoch 167/500
625/625 [==============================] - 104s 166ms/step - loss: 0.0269 - accuracy: 0.9907 - val_loss: 0.2446 - val_accuracy: 0.9333

Epoch 00167: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0167.ckpt
Epoch 168/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0282 - accuracy: 0.9898 - val_loss: 0.2266 - val_accuracy: 0.9333

Epoch 00168: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0168.ckpt
Epoch 169/500
625/625 [==============================] - 105s 169ms/step - loss: 0.0255 - accuracy: 0.9920 - val_loss: 0.4030 - val_accuracy: 0.9230

Epoch 00169: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0169.ckpt
Epoch 170/500
625/625 [==============================] - 107s 171ms/step - loss: 0.0224 - accuracy: 0.9930 - val_loss: 0.2135 - val_accuracy: 0.9427

Epoch 00170: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0170.ckpt
Epoch 171/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0276 - accuracy: 0.9907 - val_loss: 0.5118 - val_accuracy: 0.8662

Epoch 00171: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0171.ckpt
Epoch 172/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0265 - accuracy: 0.9911 - val_loss: 0.2305 - val_accuracy: 0.9475

Epoch 00172: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0172.ckpt
Epoch 173/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0224 - accuracy: 0.9909 - val_loss: 0.2885 - val_accuracy: 0.9295

Epoch 00173: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0173.ckpt
Epoch 174/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0259 - accuracy: 0.9907 - val_loss: 0.5019 - val_accuracy: 0.8685

Epoch 00174: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0174.ckpt
Epoch 175/500
625/625 [==============================] - 106s 170ms/step - loss: 0.0358 - accuracy: 0.9866 - val_loss: 0.3013 - val_accuracy: 0.9358

Epoch 00175: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0175.ckpt
Epoch 176/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0221 - accuracy: 0.9924 - val_loss: 0.3179 - val_accuracy: 0.9240

Epoch 00176: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0176.ckpt
Epoch 177/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0257 - accuracy: 0.9905 - val_loss: 0.2625 - val_accuracy: 0.9335

Epoch 00177: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0177.ckpt
Epoch 178/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0308 - accuracy: 0.9894 - val_loss: 0.3477 - val_accuracy: 0.8633

Epoch 00178: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0178.ckpt
Epoch 179/500
625/625 [==============================] - 105s 167ms/step - loss: 0.1004 - accuracy: 0.9629 - val_loss: 0.2047 - val_accuracy: 0.9400

Epoch 00179: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0179.ckpt
Epoch 180/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0320 - accuracy: 0.9890 - val_loss: 0.6238 - val_accuracy: 0.9398

Epoch 00180: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0180.ckpt
Epoch 181/500
625/625 [==============================] - 107s 171ms/step - loss: 0.0267 - accuracy: 0.9913 - val_loss: 0.2127 - val_accuracy: 0.9520

Epoch 00181: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0181.ckpt
Epoch 182/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0192 - accuracy: 0.9941 - val_loss: 0.2383 - val_accuracy: 0.9423

Epoch 00182: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0182.ckpt
Epoch 183/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0184 - accuracy: 0.9931 - val_loss: 0.3910 - val_accuracy: 0.9463

Epoch 00183: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0183.ckpt
Epoch 184/500
625/625 [==============================] - 104s 167ms/step - loss: 0.0208 - accuracy: 0.9929 - val_loss: 0.2701 - val_accuracy: 0.9390

Epoch 00184: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0184.ckpt
Epoch 185/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0216 - accuracy: 0.9928 - val_loss: 0.2123 - val_accuracy: 0.9460

Epoch 00185: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0185.ckpt
Epoch 186/500
625/625 [==============================] - 105s 168ms/step - loss: 0.0296 - accuracy: 0.9904 - val_loss: 0.3197 - val_accuracy: 0.9185

Epoch 00186: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0186.ckpt
Epoch 187/500
625/625 [==============================] - 107s 171ms/step - loss: 0.0278 - accuracy: 0.9900 - val_loss: 0.2555 - val_accuracy: 0.9153

Epoch 00187: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0187.ckpt
Epoch 188/500
625/625 [==============================] - 105s 167ms/step - loss: 0.0258 - accuracy: 0.9913 - val_loss: 0.2735 - val_accuracy: 0.9465

Epoch 00188: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0188.ckpt

In [24]:

plot_model__hist(ResNet50V2_model_adam_crawl_hist)
loss_adam,acc_adam = ResNet50V2_model_adam_crawl.evaluate(test_set, verbose=2)
print("ResNet50V2_model(Adam) 이미지 크롤링,증강 후 모델의 정확도: {:5.2f}%".format(100*acc_adam))
print("ResNet50V2_model(Adam) 이미지 크롤링,증강 후 모델의 Loss: {}".format(loss_adam))

<Figure size 432x288 with 0 Axes>

100/100 - 2s - loss: 0.2050 - accuracy: 0.9610
ResNet50V2_model(Adam) 이미지 크롤링,증강 후 모델의 정확도: 96.10%
ResNet50V2_model(Adam) 이미지 크롤링,증강 후 모델의 Loss: 0.20497991144657135

이미지 크롤링으로 가져온 데이터에 노이즈가 있어 성능이 저하될 수 있다.¶

기존의 데이터에 이미지 증강만 거쳐서 학습을 진행시켜보자.¶

1. SGD = 94.2%로 성능저하¶

In [25]:

train_set, val_set, test_set = get_dataset(aug=True,size=128)
# SGD 모델 
ResNet50V2_model_sgd_aug = keras.Sequential([
    get_ResNet50V2(size=128,trainable=True), # 백본 학습 가능하게 설정
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])

# 가중치 저장경로 변경
checkpoint_path = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path,patience=5)#동일 조건


opt = SGD(lr=0.01, decay=1e-6, momentum=0.001,nesterov=True)
ResNet50V2_model_sgd_aug.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

ResNet50V2_model_sgd_aug_hist = ResNet50V2_model_sgd_aug.fit(train_set,
                    validation_data=val_set,
                    epochs=100,
                    callbacks = [callbacks])

Found 20000 images belonging to 2 classes.
Found 4000 images belonging to 2 classes.
Found 1000 images belonging to 2 classes.
Epoch 1/100
625/625 [==============================] - 99s 152ms/step - loss: 0.5215 - accuracy: 0.7165 - val_loss: 0.2015 - val_accuracy: 0.9220

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0001.ckpt
Epoch 2/100
625/625 [==============================] - 96s 153ms/step - loss: 0.2286 - accuracy: 0.9050 - val_loss: 0.1621 - val_accuracy: 0.9352

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0002.ckpt
Epoch 3/100
625/625 [==============================] - 96s 153ms/step - loss: 0.1743 - accuracy: 0.9326 - val_loss: 0.1187 - val_accuracy: 0.9528

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0003.ckpt
Epoch 4/100
625/625 [==============================] - 96s 153ms/step - loss: 0.1478 - accuracy: 0.9438 - val_loss: 0.1145 - val_accuracy: 0.9550

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0004.ckpt
Epoch 5/100
625/625 [==============================] - 100s 160ms/step - loss: 0.1345 - accuracy: 0.9465 - val_loss: 0.1173 - val_accuracy: 0.9538

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0005.ckpt
Epoch 6/100
625/625 [==============================] - 96s 153ms/step - loss: 0.1205 - accuracy: 0.9537 - val_loss: 0.0977 - val_accuracy: 0.9603

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0006.ckpt
Epoch 7/100
625/625 [==============================] - 96s 153ms/step - loss: 0.1042 - accuracy: 0.9591 - val_loss: 0.1013 - val_accuracy: 0.9622

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0007.ckpt
Epoch 8/100
625/625 [==============================] - 96s 154ms/step - loss: 0.0893 - accuracy: 0.9652 - val_loss: 0.1345 - val_accuracy: 0.9473

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0008.ckpt
Epoch 9/100
625/625 [==============================] - 95s 153ms/step - loss: 0.0836 - accuracy: 0.9686 - val_loss: 0.1029 - val_accuracy: 0.9640

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0009.ckpt
Epoch 10/100
625/625 [==============================] - 96s 153ms/step - loss: 0.0759 - accuracy: 0.9706 - val_loss: 0.1140 - val_accuracy: 0.9610

Epoch 00010: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0010.ckpt
Epoch 11/100
625/625 [==============================] - 98s 157ms/step - loss: 0.0731 - accuracy: 0.9731 - val_loss: 0.1248 - val_accuracy: 0.9500

Epoch 00011: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0011.ckpt
Epoch 12/100
625/625 [==============================] - 96s 154ms/step - loss: 0.0683 - accuracy: 0.9732 - val_loss: 0.1319 - val_accuracy: 0.9532

Epoch 00012: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0012.ckpt
Epoch 13/100
625/625 [==============================] - 96s 153ms/step - loss: 0.0621 - accuracy: 0.9757 - val_loss: 0.1109 - val_accuracy: 0.9572

Epoch 00013: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0013.ckpt
Epoch 14/100
625/625 [==============================] - 95s 152ms/step - loss: 0.0532 - accuracy: 0.9812 - val_loss: 0.1416 - val_accuracy: 0.9523

Epoch 00014: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_aug\cp-0014.ckpt

In [26]:

plot_model__hist(ResNet50V2_model_sgd_aug_hist)
loss_sgd_aug,acc_sgd_aug = ResNet50V2_model_sgd_aug.evaluate(test_set, verbose=2)
print("ResNet50V2_model(SGD) 이미지 증강 후 모델의 정확도: {:5.2f}%".format(100*acc_sgd_aug))
print("ResNet50V2_model(SGD) 이미지 증강 후 모델의 Loss: {}".format(loss_sgd_aug))

<Figure size 432x288 with 0 Axes>

100/100 - 2s - loss: 0.1543 - accuracy: 0.9420
ResNet50V2_model(SGD) 이미지 증강 후 모델의 정확도: 94.20%
ResNet50V2_model(SGD) 이미지 증강 후 모델의 Loss: 0.15427251160144806

Adam = 94%로 성능 저하¶

In [27]:

# Adam 모델 
ResNet50V2_model_adam_aug = keras.Sequential([
    get_ResNet50V2(size=128,trainable=True), # 백본 학습 가능하게 설정
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])
# 가중치 저장경로 변경
checkpoint_path = r"D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam/cp-{epoch:04d}.ckpt"
callbacks = get_callbacks(checkpoint_path,patience=50)#동일 조건


opt = Adam(learning_rate=0.001)
ResNet50V2_model_adam_aug.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

ResNet50V2_model_adam_aug_hist = ResNet50V2_model_adam_aug.fit(train_set,
                    validation_data=val_set,
                    epochs=500,
                    callbacks = [callbacks])

Epoch 1/500
625/625 [==============================] - 99s 152ms/step - loss: 0.6789 - accuracy: 0.6105 - val_loss: 13.8396 - val_accuracy: 0.5213

Epoch 00001: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0001.ckpt
Epoch 2/500
625/625 [==============================] - 95s 152ms/step - loss: 0.5492 - accuracy: 0.7288 - val_loss: 7.0202 - val_accuracy: 0.5680

Epoch 00002: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0002.ckpt
Epoch 3/500
625/625 [==============================] - 97s 155ms/step - loss: 0.5217 - accuracy: 0.7530 - val_loss: 0.9208 - val_accuracy: 0.6313

Epoch 00003: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0003.ckpt
Epoch 4/500
625/625 [==============================] - 95s 152ms/step - loss: 0.4469 - accuracy: 0.8005 - val_loss: 0.8475 - val_accuracy: 0.7092

Epoch 00004: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0004.ckpt
Epoch 5/500
625/625 [==============================] - 95s 152ms/step - loss: 0.3896 - accuracy: 0.8332 - val_loss: 0.3060 - val_accuracy: 0.8700

Epoch 00005: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0005.ckpt
Epoch 6/500
625/625 [==============================] - 95s 151ms/step - loss: 0.3368 - accuracy: 0.8559 - val_loss: 0.6258 - val_accuracy: 0.7903

Epoch 00006: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0006.ckpt
Epoch 7/500
625/625 [==============================] - 95s 152ms/step - loss: 0.3508 - accuracy: 0.8521 - val_loss: 0.5851 - val_accuracy: 0.7055

Epoch 00007: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0007.ckpt
Epoch 8/500
625/625 [==============================] - 95s 151ms/step - loss: 0.3153 - accuracy: 0.8658 - val_loss: 0.5025 - val_accuracy: 0.7408

Epoch 00008: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0008.ckpt
Epoch 9/500
625/625 [==============================] - 98s 157ms/step - loss: 0.3448 - accuracy: 0.8564 - val_loss: 0.4017 - val_accuracy: 0.8125

Epoch 00009: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0009.ckpt
Epoch 10/500
625/625 [==============================] - 96s 153ms/step - loss: 0.2711 - accuracy: 0.8854 - val_loss: 0.3517 - val_accuracy: 0.8388

Epoch 00010: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0010.ckpt
Epoch 11/500
625/625 [==============================] - 95s 151ms/step - loss: 0.2741 - accuracy: 0.8858 - val_loss: 0.2661 - val_accuracy: 0.8900

Epoch 00011: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0011.ckpt
Epoch 12/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2697 - accuracy: 0.8905 - val_loss: 0.2745 - val_accuracy: 0.8815

Epoch 00012: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0012.ckpt
Epoch 13/500
625/625 [==============================] - 95s 151ms/step - loss: 0.2429 - accuracy: 0.8985 - val_loss: 0.5356 - val_accuracy: 0.7853

Epoch 00013: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0013.ckpt
Epoch 14/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2283 - accuracy: 0.9077 - val_loss: 0.2876 - val_accuracy: 0.8813

Epoch 00014: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0014.ckpt
Epoch 15/500
625/625 [==============================] - 97s 154ms/step - loss: 0.2241 - accuracy: 0.9091 - val_loss: 0.2593 - val_accuracy: 0.8888

Epoch 00015: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0015.ckpt
Epoch 16/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2103 - accuracy: 0.9143 - val_loss: 0.3032 - val_accuracy: 0.8635

Epoch 00016: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0016.ckpt
Epoch 17/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2527 - accuracy: 0.9006 - val_loss: 0.2277 - val_accuracy: 0.8988

Epoch 00017: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0017.ckpt
Epoch 18/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2186 - accuracy: 0.9115 - val_loss: 0.3183 - val_accuracy: 0.8505

Epoch 00018: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0018.ckpt
Epoch 19/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2083 - accuracy: 0.9138 - val_loss: 0.1932 - val_accuracy: 0.9172

Epoch 00019: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0019.ckpt
Epoch 20/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2141 - accuracy: 0.9144 - val_loss: 0.5673 - val_accuracy: 0.6975

Epoch 00020: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0020.ckpt
Epoch 21/500
625/625 [==============================] - 95s 152ms/step - loss: 0.2128 - accuracy: 0.9120 - val_loss: 0.2766 - val_accuracy: 0.8960

Epoch 00021: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0021.ckpt
Epoch 22/500
625/625 [==============================] - 98s 156ms/step - loss: 0.1808 - accuracy: 0.9272 - val_loss: 0.3229 - val_accuracy: 0.8612

Epoch 00022: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0022.ckpt
Epoch 23/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1749 - accuracy: 0.9289 - val_loss: 0.7760 - val_accuracy: 0.7032

Epoch 00023: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0023.ckpt
Epoch 24/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1776 - accuracy: 0.9304 - val_loss: 0.3591 - val_accuracy: 0.8585

Epoch 00024: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0024.ckpt
Epoch 25/500
625/625 [==============================] - 95s 151ms/step - loss: 0.1628 - accuracy: 0.9331 - val_loss: 0.4409 - val_accuracy: 0.8123

Epoch 00025: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0025.ckpt
Epoch 26/500
625/625 [==============================] - 95s 151ms/step - loss: 0.2411 - accuracy: 0.9035 - val_loss: 0.2518 - val_accuracy: 0.8865

Epoch 00026: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0026.ckpt
Epoch 27/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1962 - accuracy: 0.9215 - val_loss: 0.2472 - val_accuracy: 0.8935

Epoch 00027: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0027.ckpt
Epoch 28/500
625/625 [==============================] - 98s 156ms/step - loss: 0.1549 - accuracy: 0.9393 - val_loss: 0.4941 - val_accuracy: 0.7818

Epoch 00028: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0028.ckpt
Epoch 29/500
625/625 [==============================] - 97s 156ms/step - loss: 0.1733 - accuracy: 0.9333 - val_loss: 0.1753 - val_accuracy: 0.9323

Epoch 00029: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0029.ckpt
Epoch 30/500
625/625 [==============================] - 94s 151ms/step - loss: 0.1502 - accuracy: 0.9385 - val_loss: 0.1724 - val_accuracy: 0.9285

Epoch 00030: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0030.ckpt
Epoch 31/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1484 - accuracy: 0.9433 - val_loss: 0.1920 - val_accuracy: 0.9287

Epoch 00031: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0031.ckpt
Epoch 32/500
625/625 [==============================] - 95s 151ms/step - loss: 0.1501 - accuracy: 0.9396 - val_loss: 0.5099 - val_accuracy: 0.8202

Epoch 00032: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0032.ckpt
Epoch 33/500
625/625 [==============================] - 95s 151ms/step - loss: 0.1544 - accuracy: 0.9357 - val_loss: 0.2111 - val_accuracy: 0.9153

Epoch 00033: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0033.ckpt
Epoch 34/500
625/625 [==============================] - 97s 155ms/step - loss: 0.1358 - accuracy: 0.9460 - val_loss: 0.1870 - val_accuracy: 0.9208

Epoch 00034: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0034.ckpt
Epoch 35/500
625/625 [==============================] - 95s 153ms/step - loss: 0.1363 - accuracy: 0.9478 - val_loss: 0.2436 - val_accuracy: 0.9062

Epoch 00035: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0035.ckpt
Epoch 36/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1360 - accuracy: 0.9473 - val_loss: 0.1595 - val_accuracy: 0.9330

Epoch 00036: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0036.ckpt
Epoch 37/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1290 - accuracy: 0.9482 - val_loss: 0.1550 - val_accuracy: 0.9380

Epoch 00037: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0037.ckpt
Epoch 38/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1242 - accuracy: 0.9511 - val_loss: 0.1919 - val_accuracy: 0.9165

Epoch 00038: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0038.ckpt
Epoch 39/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1219 - accuracy: 0.9543 - val_loss: 0.1999 - val_accuracy: 0.9168

Epoch 00039: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0039.ckpt
Epoch 40/500
625/625 [==============================] - 98s 156ms/step - loss: 0.1747 - accuracy: 0.9322 - val_loss: 0.1791 - val_accuracy: 0.9252

Epoch 00040: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0040.ckpt
Epoch 41/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1309 - accuracy: 0.9497 - val_loss: 0.2490 - val_accuracy: 0.8940

Epoch 00041: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0041.ckpt
Epoch 42/500
625/625 [==============================] - 95s 153ms/step - loss: 0.2099 - accuracy: 0.9164 - val_loss: 0.2947 - val_accuracy: 0.8790

Epoch 00042: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0042.ckpt
Epoch 43/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1785 - accuracy: 0.9285 - val_loss: 0.1518 - val_accuracy: 0.9367

Epoch 00043: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0043.ckpt
Epoch 44/500
625/625 [==============================] - 95s 151ms/step - loss: 0.1281 - accuracy: 0.9487 - val_loss: 0.1881 - val_accuracy: 0.9205

Epoch 00044: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0044.ckpt
Epoch 45/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1173 - accuracy: 0.9542 - val_loss: 0.1499 - val_accuracy: 0.9423

Epoch 00045: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0045.ckpt
Epoch 46/500
625/625 [==============================] - 96s 154ms/step - loss: 0.1091 - accuracy: 0.9581 - val_loss: 0.1740 - val_accuracy: 0.9373

Epoch 00046: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0046.ckpt
Epoch 47/500
625/625 [==============================] - 96s 154ms/step - loss: 0.1065 - accuracy: 0.9584 - val_loss: 0.1693 - val_accuracy: 0.9375

Epoch 00047: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0047.ckpt
Epoch 48/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1092 - accuracy: 0.9571 - val_loss: 0.1878 - val_accuracy: 0.9333

Epoch 00048: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0048.ckpt
Epoch 49/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1055 - accuracy: 0.9604 - val_loss: 0.1727 - val_accuracy: 0.9317

Epoch 00049: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0049.ckpt
Epoch 50/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1078 - accuracy: 0.9588 - val_loss: 0.1579 - val_accuracy: 0.9330

Epoch 00050: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0050.ckpt
Epoch 51/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1029 - accuracy: 0.9600 - val_loss: 0.1902 - val_accuracy: 0.9183

Epoch 00051: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0051.ckpt
Epoch 52/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1052 - accuracy: 0.9602 - val_loss: 0.1569 - val_accuracy: 0.9410

Epoch 00052: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0052.ckpt
Epoch 53/500
625/625 [==============================] - 97s 156ms/step - loss: 0.1019 - accuracy: 0.9618 - val_loss: 0.2001 - val_accuracy: 0.9205

Epoch 00053: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0053.ckpt
Epoch 54/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1088 - accuracy: 0.9579 - val_loss: 0.1972 - val_accuracy: 0.9220

Epoch 00054: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0054.ckpt
Epoch 55/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0994 - accuracy: 0.9586 - val_loss: 0.2066 - val_accuracy: 0.9170

Epoch 00055: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0055.ckpt
Epoch 56/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1057 - accuracy: 0.9606 - val_loss: 0.1782 - val_accuracy: 0.9287

Epoch 00056: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0056.ckpt
Epoch 57/500
625/625 [==============================] - 95s 153ms/step - loss: 0.0876 - accuracy: 0.9657 - val_loss: 0.1793 - val_accuracy: 0.9222

Epoch 00057: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0057.ckpt
Epoch 58/500
625/625 [==============================] - 95s 151ms/step - loss: 0.0907 - accuracy: 0.9650 - val_loss: 0.1720 - val_accuracy: 0.9350

Epoch 00058: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0058.ckpt
Epoch 59/500
625/625 [==============================] - 97s 156ms/step - loss: 0.1210 - accuracy: 0.9556 - val_loss: 0.1775 - val_accuracy: 0.9352

Epoch 00059: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0059.ckpt
Epoch 60/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0885 - accuracy: 0.9666 - val_loss: 0.1713 - val_accuracy: 0.9342

Epoch 00060: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0060.ckpt
Epoch 61/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0864 - accuracy: 0.9656 - val_loss: 0.1683 - val_accuracy: 0.9310

Epoch 00061: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0061.ckpt
Epoch 62/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0864 - accuracy: 0.9679 - val_loss: 0.1734 - val_accuracy: 0.9310

Epoch 00062: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0062.ckpt
Epoch 63/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0905 - accuracy: 0.9647 - val_loss: 0.1851 - val_accuracy: 0.9367

Epoch 00063: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0063.ckpt
Epoch 64/500
625/625 [==============================] - 95s 152ms/step - loss: 0.1040 - accuracy: 0.9622 - val_loss: 0.2076 - val_accuracy: 0.9227

Epoch 00064: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0064.ckpt
Epoch 65/500
625/625 [==============================] - 98s 157ms/step - loss: 0.0839 - accuracy: 0.9687 - val_loss: 0.1538 - val_accuracy: 0.9430

Epoch 00065: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0065.ckpt
Epoch 66/500
625/625 [==============================] - 98s 156ms/step - loss: 0.0804 - accuracy: 0.9701 - val_loss: 0.1949 - val_accuracy: 0.9252

Epoch 00066: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0066.ckpt
Epoch 67/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0796 - accuracy: 0.9683 - val_loss: 0.2779 - val_accuracy: 0.9040

Epoch 00067: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0067.ckpt
Epoch 68/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0774 - accuracy: 0.9710 - val_loss: 0.1928 - val_accuracy: 0.9250

Epoch 00068: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0068.ckpt
Epoch 69/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0770 - accuracy: 0.9703 - val_loss: 0.1993 - val_accuracy: 0.9243

Epoch 00069: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0069.ckpt
Epoch 70/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0740 - accuracy: 0.9722 - val_loss: 7.0663 - val_accuracy: 0.6507

Epoch 00070: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0070.ckpt
Epoch 71/500
625/625 [==============================] - 97s 156ms/step - loss: 0.1479 - accuracy: 0.9413 - val_loss: 0.1738 - val_accuracy: 0.9415

Epoch 00071: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0071.ckpt
Epoch 72/500
625/625 [==============================] - 96s 153ms/step - loss: 0.0792 - accuracy: 0.9718 - val_loss: 0.1860 - val_accuracy: 0.9258

Epoch 00072: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0072.ckpt
Epoch 73/500
625/625 [==============================] - 95s 153ms/step - loss: 0.0742 - accuracy: 0.9718 - val_loss: 0.1429 - val_accuracy: 0.9467

Epoch 00073: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0073.ckpt
Epoch 74/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0636 - accuracy: 0.9786 - val_loss: 0.2485 - val_accuracy: 0.9193

Epoch 00074: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0074.ckpt
Epoch 75/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0682 - accuracy: 0.9751 - val_loss: 0.1444 - val_accuracy: 0.9517

Epoch 00075: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0075.ckpt
Epoch 76/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0680 - accuracy: 0.9754 - val_loss: 0.1696 - val_accuracy: 0.9450

Epoch 00076: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0076.ckpt
Epoch 77/500
625/625 [==============================] - 96s 154ms/step - loss: 0.0728 - accuracy: 0.9746 - val_loss: 0.1712 - val_accuracy: 0.9342

Epoch 00077: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0077.ckpt
Epoch 78/500
625/625 [==============================] - 98s 156ms/step - loss: 0.0686 - accuracy: 0.9752 - val_loss: 0.1518 - val_accuracy: 0.9463

Epoch 00078: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0078.ckpt
Epoch 79/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0634 - accuracy: 0.9757 - val_loss: 0.2237 - val_accuracy: 0.9047

Epoch 00079: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0079.ckpt
Epoch 80/500
625/625 [==============================] - 96s 153ms/step - loss: 0.0659 - accuracy: 0.9747 - val_loss: 0.2143 - val_accuracy: 0.9277

Epoch 00080: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0080.ckpt
Epoch 81/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0617 - accuracy: 0.9772 - val_loss: 0.2300 - val_accuracy: 0.9143

Epoch 00081: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0081.ckpt
Epoch 82/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0693 - accuracy: 0.9743 - val_loss: 0.1865 - val_accuracy: 0.9302

Epoch 00082: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0082.ckpt
Epoch 83/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0601 - accuracy: 0.9770 - val_loss: 0.1918 - val_accuracy: 0.9317

Epoch 00083: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0083.ckpt
Epoch 84/500
625/625 [==============================] - 101s 161ms/step - loss: 0.0622 - accuracy: 0.9769 - val_loss: 0.1557 - val_accuracy: 0.9445

Epoch 00084: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0084.ckpt
Epoch 85/500
625/625 [==============================] - 99s 158ms/step - loss: 0.0598 - accuracy: 0.9773 - val_loss: 0.3101 - val_accuracy: 0.8988

Epoch 00085: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0085.ckpt
Epoch 86/500
625/625 [==============================] - 96s 154ms/step - loss: 0.0624 - accuracy: 0.9757 - val_loss: 0.2430 - val_accuracy: 0.9245

Epoch 00086: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0086.ckpt
Epoch 87/500
625/625 [==============================] - 96s 153ms/step - loss: 0.0593 - accuracy: 0.9766 - val_loss: 0.2724 - val_accuracy: 0.9128

Epoch 00087: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0087.ckpt
Epoch 88/500
625/625 [==============================] - 96s 154ms/step - loss: 0.0649 - accuracy: 0.9764 - val_loss: 0.1770 - val_accuracy: 0.9333

Epoch 00088: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0088.ckpt
Epoch 89/500
625/625 [==============================] - 97s 155ms/step - loss: 0.0584 - accuracy: 0.9791 - val_loss: 0.2046 - val_accuracy: 0.9367

Epoch 00089: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0089.ckpt
Epoch 90/500
625/625 [==============================] - 106s 169ms/step - loss: 0.0577 - accuracy: 0.9796 - val_loss: 0.2203 - val_accuracy: 0.9202

Epoch 00090: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0090.ckpt
Epoch 91/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0630 - accuracy: 0.9771 - val_loss: 0.2010 - val_accuracy: 0.9355

Epoch 00091: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0091.ckpt
Epoch 92/500
625/625 [==============================] - 95s 151ms/step - loss: 0.0508 - accuracy: 0.9789 - val_loss: 0.1837 - val_accuracy: 0.9385

Epoch 00092: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0092.ckpt
Epoch 93/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0530 - accuracy: 0.9793 - val_loss: 0.1845 - val_accuracy: 0.9435

Epoch 00093: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0093.ckpt
Epoch 94/500
625/625 [==============================] - 95s 151ms/step - loss: 0.0511 - accuracy: 0.9814 - val_loss: 0.2883 - val_accuracy: 0.8997

Epoch 00094: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0094.ckpt
Epoch 95/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0492 - accuracy: 0.9818 - val_loss: 0.1858 - val_accuracy: 0.9317

Epoch 00095: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0095.ckpt
Epoch 96/500
625/625 [==============================] - 98s 156ms/step - loss: 0.0569 - accuracy: 0.9785 - val_loss: 0.2229 - val_accuracy: 0.9283

Epoch 00096: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0096.ckpt
Epoch 97/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0498 - accuracy: 0.9823 - val_loss: 0.1728 - val_accuracy: 0.9373

Epoch 00097: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0097.ckpt
Epoch 98/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0472 - accuracy: 0.9825 - val_loss: 0.2010 - val_accuracy: 0.9370

Epoch 00098: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0098.ckpt
Epoch 99/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0571 - accuracy: 0.9801 - val_loss: 0.2070 - val_accuracy: 0.9222

Epoch 00099: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0099.ckpt
Epoch 100/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0498 - accuracy: 0.9828 - val_loss: 0.2027 - val_accuracy: 0.9452

Epoch 00100: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0100.ckpt
Epoch 101/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0494 - accuracy: 0.9803 - val_loss: 0.1947 - val_accuracy: 0.9243

Epoch 00101: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0101.ckpt
Epoch 102/500
625/625 [==============================] - 97s 155ms/step - loss: 0.0520 - accuracy: 0.9810 - val_loss: 0.1659 - val_accuracy: 0.9435

Epoch 00102: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0102.ckpt
Epoch 103/500
625/625 [==============================] - 98s 156ms/step - loss: 0.0467 - accuracy: 0.9834 - val_loss: 0.3354 - val_accuracy: 0.8978

Epoch 00103: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0103.ckpt
Epoch 104/500
625/625 [==============================] - 96s 153ms/step - loss: 0.0508 - accuracy: 0.9812 - val_loss: 0.5619 - val_accuracy: 0.8360

Epoch 00104: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0104.ckpt
Epoch 105/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0493 - accuracy: 0.9832 - val_loss: 0.3258 - val_accuracy: 0.8650

Epoch 00105: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0105.ckpt
Epoch 106/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0466 - accuracy: 0.9829 - val_loss: 0.2874 - val_accuracy: 0.9273

Epoch 00106: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0106.ckpt
Epoch 107/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0481 - accuracy: 0.9830 - val_loss: 0.1911 - val_accuracy: 0.9355

Epoch 00107: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0107.ckpt
Epoch 108/500
625/625 [==============================] - 96s 154ms/step - loss: 0.0463 - accuracy: 0.9820 - val_loss: 0.1735 - val_accuracy: 0.9485

Epoch 00108: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0108.ckpt
Epoch 109/500
625/625 [==============================] - 96s 153ms/step - loss: 0.0460 - accuracy: 0.9841 - val_loss: 0.2539 - val_accuracy: 0.9185

Epoch 00109: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0109.ckpt
Epoch 110/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0454 - accuracy: 0.9823 - val_loss: 0.2024 - val_accuracy: 0.9345

Epoch 00110: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0110.ckpt
Epoch 111/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0396 - accuracy: 0.9858 - val_loss: 0.1786 - val_accuracy: 0.9442

Epoch 00111: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0111.ckpt
Epoch 112/500
625/625 [==============================] - 94s 151ms/step - loss: 0.0432 - accuracy: 0.9844 - val_loss: 0.1809 - val_accuracy: 0.9377

Epoch 00112: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0112.ckpt
Epoch 113/500
625/625 [==============================] - 95s 151ms/step - loss: 0.0427 - accuracy: 0.9843 - val_loss: 0.2156 - val_accuracy: 0.9330

Epoch 00113: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0113.ckpt
Epoch 114/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0432 - accuracy: 0.9845 - val_loss: 0.1710 - val_accuracy: 0.9442

Epoch 00114: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0114.ckpt
Epoch 115/500
625/625 [==============================] - 98s 156ms/step - loss: 0.0377 - accuracy: 0.9873 - val_loss: 0.1644 - val_accuracy: 0.9448

Epoch 00115: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0115.ckpt
Epoch 116/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0423 - accuracy: 0.9845 - val_loss: 0.2065 - val_accuracy: 0.9402

Epoch 00116: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0116.ckpt
Epoch 117/500
625/625 [==============================] - 95s 153ms/step - loss: 0.0386 - accuracy: 0.9862 - val_loss: 0.1610 - val_accuracy: 0.9452

Epoch 00117: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0117.ckpt
Epoch 118/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0366 - accuracy: 0.9852 - val_loss: 0.3156 - val_accuracy: 0.9047

Epoch 00118: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0118.ckpt
Epoch 119/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0436 - accuracy: 0.9840 - val_loss: 0.1731 - val_accuracy: 0.9475

Epoch 00119: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0119.ckpt
Epoch 120/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0441 - accuracy: 0.9839 - val_loss: 0.1845 - val_accuracy: 0.9425

Epoch 00120: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0120.ckpt
Epoch 121/500
625/625 [==============================] - 97s 155ms/step - loss: 0.0399 - accuracy: 0.9848 - val_loss: 0.2213 - val_accuracy: 0.9340

Epoch 00121: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0121.ckpt
Epoch 122/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0345 - accuracy: 0.9879 - val_loss: 0.1736 - val_accuracy: 0.9450

Epoch 00122: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0122.ckpt
Epoch 123/500
625/625 [==============================] - 96s 153ms/step - loss: 0.0329 - accuracy: 0.9888 - val_loss: 0.2234 - val_accuracy: 0.9227

Epoch 00123: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0123.ckpt
Epoch 124/500
625/625 [==============================] - 95s 151ms/step - loss: 0.0434 - accuracy: 0.9839 - val_loss: 0.2355 - val_accuracy: 0.9230

Epoch 00124: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0124.ckpt
Epoch 125/500
625/625 [==============================] - 95s 152ms/step - loss: 0.0359 - accuracy: 0.9876 - val_loss: 0.2075 - val_accuracy: 0.9360

Epoch 00125: saving model to D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_Adam\cp-0125.ckpt

In [28]:

plot_model__hist(ResNet50V2_model_adam_aug_hist)
loss_adam_aug,acc_adam_aug = ResNet50V2_model_adam_aug.evaluate(test_set, verbose=2)
print("ResNet50V2_model(Adam) 이미지 증강 후 모델의 정확도: {:5.2f}%".format(100*acc_adam_aug))
print("ResNet50V2_model(Adam) 이미지 증강 후 모델의 Loss: {}".format(loss_adam_aug))

<Figure size 432x288 with 0 Axes>

100/100 - 2s - loss: 0.1966 - accuracy: 0.9400
ResNet50V2_model(Adam) 이미지 증강 후 모델의 정확도: 94.00%
ResNet50V2_model(Adam) 이미지 증강 후 모델의 Loss: 0.19657617807388306

<최종 결론>¶

• SGD : 97%

• Adam : 82.4%

• Crawling, Augmented SGD : 97.5% =======>최고 성능

• Crawling, Augmented Adam : 96.1%

• Augmented SGD : 94.2%

• Augmented Adam : 94%

최고성능 도출한 모델 가중치 복구하여 이미지 분류 해보기¶

In [21]:

# 모델 생성해서 가중치 복구하여 임의의 이미지에 분류 넣기

# SGD 모델 
from tensorflow.keras.optimizers import SGD
model = Sequential([
    get_ResNet50V2(size=128,trainable=True), # 백본 학습 가능하게 설정
    Dense(256, activation='relu'),
    Dropout(0.5),
    Dense(128, activation='relu'),
    Dropout(0.25),
    Dense(1,activation='sigmoid')
])


opt = SGD(lr=0.01, decay=1e-6, momentum=0.001,nesterov=True)
model.compile(optimizer=opt, loss='binary_crossentropy',metrics=['accuracy'])

model.load_weights(r'D:\Project(2)\cat&dog_set\check_point_ResNet50V2_crawl_SGD\cp-0020.ckpt')

Out[21]:

<tensorflow.python.training.tracking.util.CheckpointLoadStatus at 0x137f1371a88>

In [60]:

loss,acc = model.evaluate(test_set, verbose=2)
print("ResNet50V2 최종 모델의 정확도: {:5.2f}%".format(100*acc))
print("ResNet50V2 최종 모델의 Loss: {}".format(loss))

100/100 - 2s - loss: 0.0836 - accuracy: 0.9750
ResNet50V2 최종 모델의 정확도: 97.50%
ResNet50V2 최종 모델의 Loss: 0.083615742623806

In [90]:

import cv2
import matplotlib.pyplot as plt
import tensorflow as tf

test_path = r"C:\Users\cvLab\Desktop\test"
test = os.listdir(test_path)
test_sample = []
for i in range(9):
    test_sample.append(os.path.join(test_path,test[i]))

images = [cv2.imread(i) for i in test_sample]
for i,j in enumerate(images):
    images[i] = cv2.resize(j,(128,128), interpolation=cv2.INTER_AREA)
idx=1
plt.figure(figsize=(16,16))

category = {0:'cat', 1:'dog'}
for i,title, test_img in zip(images,test,test_sample):
    plt.subplot(3,3,idx)
    plt.title(title)
    plt.imshow(cv2.cvtColor(i,cv2.COLOR_BGR2RGB))
    test_image = load_img(test_img)
    test = tf.keras.preprocessing.image.img_to_array(test_image,data_format=None, dtype=None)
    test = test.reshape((1,)+test.shape)
    result = model.predict(test/255.)
    plt.xlabel(category[int(round(result[0][0],0))])
    # print(int(round(result[0][0],0)))
    idx+=1