CNN(Convolutional Neural Network)-이미지 데이터 셋을 이용한 CNN Modeling

Dogs and Cats Image_Data

• Train_Data : 2000(1000_Dogs, 1000_Cats)
• Valid_Data : 1000(500_Dogs, 500_Cats)
• Test_Data : 1000(500_Dogs, 500_Cats)

'dogs_and_cats_small.zip' 디렉토리를 구글드라이브에 업로드

from google.colab import drive

drive.mount('/content/drive')

---

1. Data Preprocessing

import tensorflow as tf

!unzip /content/drive/My\ Drive/Colab\ Notebooks/datasets/dogs_and_cats_small.zip

1-1. Image_File Directory Setting

train_dir = 'train'
valid_dir = 'validation'
test_dir  = 'test'

---

1-2. ImageDataGenerator( ) & flow_from_directory( )

from tensorflow.keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(rescale = 1./255)
valid_datagen = ImageDataGenerator(rescale = 1./255)

train_generator = train_datagen.flow_from_directory(
                  train_dir,
                  target_size = (150, 150),
                  batch_size = 20,
                  class_mode = 'binary')

valid_generator = valid_datagen.flow_from_directory(
                  valid_dir,
                  target_size = (150, 150),
                  batch_size = 20,
                  class_mode = 'binary')

Found 2000 images belonging to 2 classes.

Found 1000 images belonging to 2 classes.

---

1-3. Test train_generator

for data_batch, labels_batch in train_generator:
    print('배치 데이터 크기:', data_batch.shape)
    print('배치 레이블 크기:', labels_batch.shape)
    break

labels_batch

---

2. CNN Keras Modeling

2-1.  Model Define

#Feature Extraction & Classification

from tensorflow.keras import layers
from tensorflow.keras import models

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation = 'relu', input_shape = (150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation = 'relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation = 'relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation = 'relu'))
model.add(layers.MaxPooling2D((2, 2)))

model.add(layers.Flatten())
model.add(layers.Dense(512, activation = 'relu'))
model.add(layers.Dense(1, activation = 'sigmoid'))

model.summary()

---

2-2. Model Compile

model.compile(loss = 'binary_crossentropy',
              optimizer = 'adam',
              metrics = ['accuracy'])

---

2-3. Model Fit

%%time

Hist_dandc = model.fit(train_generator,
                       steps_per_epoch = 100,
                       epochs = 60,
                       validation_data = valid_generator,
                       validation_steps = 50)

---

2-4. 학습 결과 시각화

#Loss Visualization

import matplotlib.pyplot as plt

epochs = range(1, len(Hist_dandc.history['loss']) + 1)

plt.figure(figsize = (9, 6))
plt.plot(epochs, Hist_dandc.history['loss'])
plt.plot(epochs, Hist_dandc.history['val_loss'])

plt.title('Training & Validation Loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.grid()
plt.show()

#Accuracy Visualization

import matplotlib.pyplot as plt

epochs = range(1, len(Hist_dandc.history['loss']) + 1)

plt.figure(figsize = (9, 6))
plt.plot(epochs, Hist_dandc.history['accuracy'])
plt.plot(epochs, Hist_dandc.history['val_accuracy'])

plt.title('Training & Validation Accuracy')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend(['Training Accuracy', 'Validation Accuracy'])
plt.grid()
plt.show()

---

2-5. Model Evaluate

test_datagen = ImageDataGenerator(rescale = 1./255)

test_generator = test_datagen.flow_from_directory(
                 test_dir,
                 target_size = (150, 150),
                 batch_size = 20,
                 class_mode = 'binary')

Found 1000 images belonging to 2 classes.

loss, accuracy = model.evaluate(test_generator,
                                steps = 50)

print('Loss = {:.5f}'.format(loss))
print('Accuracy = {:.5f}'.format(accuracy))

---

3. Model Save

model.save('/content/drive/My Drive/Colab Notebooks/models/002_dogs_and_cats_small.h5')

---

Overfitting 대응책

[Optional] Image Augmentation Test

• rotation_range = 40 : 0도에서 40도 사이에서 임의의 각도록 회전
• width_shift_range = 0.2 : 20% 픽셀 내외로 좌우 이동
• height_shift_range = 0.2 : 20% 픽셀 내외로 상하 이동
• shear_range = 0.2 : 0.2 라디안 내외로 시계 반대방향으로 변형
• zoom_range = 0.2 : 80%에서 120% 범위에서 확대/축소
• horizontal_flip = True : 수평방향 뒤집기
• vertical_flip = True : 수직방향 뒤집기
• fill_mode = 'nearest' : 주변 픽셀로 이미지 채우기

from tensorflow.keras.preprocessing.image import ImageDataGenerator

datagen = ImageDataGenerator(rotation_range = 40,
                             width_shift_range = 0.2,
                             height_shift_range = 0.2,
                             shear_range = 0.2,
                             zoom_range = 0.2,
                             horizontal_flip = True,
                             vertical_flip = True,
                             fill_mode = 'nearest')

from tensorflow.keras.preprocessing import image
import matplotlib.pyplot as plt
import os

train_cats_dir = train_dir = os.path.join('train', 'cats')
fnames = sorted([os.path.join(train_cats_dir, fname) for fname in os.listdir(train_cats_dir)])

# 테스트 이미지 선택
img_path = fnames[77]

# 이미지 읽고 크기 변경
img = image.load_img(img_path, target_size=(150, 150))

# (150, 150, 3) 배열 변환
x = image.img_to_array(img)

# (1, 150, 150, 3) 변환
x = x.reshape((1,) + x.shape)

# 랜덤하게 변환된 이미지 배치 생성
i = 0

for batch in datagen.flow(x, batch_size=1):
    plt.figure(i)
    imgplot = plt.imshow(image.array_to_img(batch[0]))
    i += 1
    if i % 4 == 0:
        break

plt.show()

---

train_dir = 'train'
valid_dir = 'validation'
test_dir = 'test'

from tensorflow.keras.preprocessing.image import ImageDataGenerator

# With Augmentation
train_datagen = ImageDataGenerator(rescale = 1./255,
                                   rotation_range = 40,
                                   width_shift_range = 0.2,
                                   height_shift_range = 0.2,
                                   shear_range = 0.2,
                                   zoom_range = 0.2,
                                   horizontal_flip=True,
                                   vertical_flip = True,
                                   fill_mode = 'nearest')

# Without Augmentation 
valid_datagen = ImageDataGenerator(rescale = 1./255)



# With Augmentation
train_generator = train_datagen.flow_from_directory(
                  train_dir,
                  target_size = (150, 150),
                  batch_size = 20,
                  class_mode = 'binary')

# Without Augmentation
valid_generator = valid_datagen.flow_from_directory(
                  valid_dir,
                  target_size = (150, 150),
                  batch_size = 20,
                  class_mode = 'binary')

from tensorflow.keras import layers
from tensorflow.keras import models

model = models.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation = 'relu', input_shape = (150, 150, 3)))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation = 'relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation = 'relu'))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(128, (3, 3), activation = 'relu'))
model.add(layers.MaxPooling2D((2, 2)))

model.add(layers.Flatten())
model.add(layers.Dropout(0.5))
model.add(layers.Dense(512, activation = 'relu'))
model.add(layers.Dense(1, activation = 'sigmoid'))

model.summary()

model.compile(loss = 'binary_crossentropy',
              optimizer = 'adam',
              metrics = ['accuracy'])

%%time
#epochs : 60 -> 100

Hist_dandc = model.fit(train_generator,
                       steps_per_epoch = 100,            
                       epochs = 100,
                       validation_data = valid_generator,
                       validation_steps = 50)

import matplotlib.pyplot as plt

epochs = range(1, len(Hist_dandc.history['loss']) + 1)

plt.figure(figsize = (9, 6))
plt.plot(epochs, Hist_dandc.history['loss'])
plt.plot(epochs, Hist_dandc.history['val_loss'])

plt.title('Training & Validation Loss')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend(['Training Loss', 'Validation Loss'])
plt.grid()
plt.show()

import matplotlib.pyplot as plt

epochs = range(1, len(Hist_dandc.history['loss']) + 1)

plt.figure(figsize = (9, 6))
plt.plot(epochs, Hist_dandc.history['accuracy'])
plt.plot(epochs, Hist_dandc.history['val_accuracy'])

plt.title('Training & Validation Accuracy')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend(['Training Accuracy', 'Validation Accuracy'])
plt.grid()
plt.show()

test_datagen = ImageDataGenerator(rescale = 1./255)

test_generator = test_datagen.flow_from_directory(
                 test_dir,
                 target_size = (150, 150),
                 batch_size = 20,
                 class_mode = 'binary')

loss, accuracy = model.evaluate(test_generator,
                                steps = 50)

print('Loss = {:.5f}'.format(loss))
print('Accuracy = {:.5f}'.format(accuracy))

model.save('/content/drive/My Drive/Colab Notebooks/models/003_dogs_and_cats_augmentation.h5')