[Generative Model] 단일 이미지 오토인코더

라이브러리

 

from matplotlib.pyplot import imshow
import numpy as np
import cv2
from tensorflow.keras.utils import img_to_array
from tensorflow.keras.layers import Conv2D, MaxPooling2D, UpSampling2D
from tensorflow.keras.models import Sequential

np.random.seed(42)

SIZE = 256
img_data = []

img = cv2.imread('monalisa.jpg', 1)   # Change 1 to 0 for grey images
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB) # 이미지를 BGR에서 RGB로 변환
img = cv2.resize(img,(SIZE, SIZE))
img_data.append(img_to_array(img))

img_array = np.reshape(img_data, (len(img_data), SIZE, SIZE, 3))
img_array = img_array.astype('float32') / 255.

 

모델 구성

 

model = Sequential()

model.add(Conv2D(32, (3, 3), activation='relu', padding='same', input_shape=(SIZE, SIZE, 3)))
model.add(MaxPooling2D((2, 2), padding='same'))
model.add(Conv2D(8, (3, 3), activation='relu', padding='same'))
model.add(MaxPooling2D((2, 2), padding='same'))
model.add(Conv2D(8, (3, 3), activation='relu', padding='same'))
model.add(MaxPooling2D((2, 2), padding='same'))
model.add(Conv2D(8, (3, 3), activation='relu', padding='same'))
model.add(UpSampling2D((2, 2)))
model.add(Conv2D(8, (3, 3), activation='relu', padding='same'))
model.add(UpSampling2D((2, 2)))
model.add(Conv2D(32, (3, 3), activation='relu', padding='same'))
model.add(UpSampling2D((2, 2)))
model.add(Conv2D(3, (3, 3), activation='relu', padding='same'))

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

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d (Conv2D)             (None, 256, 256, 32)      896       
                                                                 
 max_pooling2d (MaxPooling2D  (None, 128, 128, 32)     0         
 )                                                               
                                                                 
 conv2d_1 (Conv2D)           (None, 128, 128, 8)       2312      
                                                                 
 max_pooling2d_1 (MaxPooling  (None, 64, 64, 8)        0         
 2D)                                                             
                                                                 
 conv2d_2 (Conv2D)           (None, 64, 64, 8)         584       
                                                                 
 max_pooling2d_2 (MaxPooling  (None, 32, 32, 8)        0         
 2D)                                                             
                                                                 
 conv2d_3 (Conv2D)           (None, 32, 32, 8)         584       
                                                                 
 up_sampling2d (UpSampling2D  (None, 64, 64, 8)        0         
 )                                                               
                                                                 
 conv2d_4 (Conv2D)           (None, 64, 64, 8)         584       
                                                                 
 up_sampling2d_1 (UpSampling  (None, 128, 128, 8)      0         
 2D)                                                             
                                                                 
 conv2d_5 (Conv2D)           (None, 128, 128, 32)      2336      
                                                                 
 up_sampling2d_2 (UpSampling  (None, 256, 256, 32)     0         
 2D)                                                             
                                                                 
 conv2d_6 (Conv2D)           (None, 256, 256, 3)       867       
                                                                 
=================================================================
Total params: 8,163
Trainable params: 8,163
Non-trainable params: 0
_________________________________________________________________

 

모델 학습

 

model.fit(img_array, img_array,
        epochs = 5000,
        shuffle = True)

Epoch 5000/5000
1/1 [==============================] - 0s 286ms/step - loss: 0.0013 - accuracy: 0.9187

pred = model.predict(img_array)

imshow(pred[0].reshape(SIZE, SIZE, 3), cmap = "gray")