TensorFlow

• Custom Dataset 만들기 (from generator)

class DataGenerator(tf.keras.utils.Sequence):
    'Generates data for Keras'
    def __init__(self, df, img_size=256, batch_size=32, path=BASE): 
        self.df = df
        self.img_size = img_size
        self.batch_size = batch_size
        self.path = path
        self.indexes = np.arange( len(self.df) )
        
    def __len__(self):
        'Denotes the number of batches per epoch'
        ct = len(self.df) // self.batch_size
        ct += int(( (len(self.df)) % self.batch_size)!=0)
        return ct
 
    def __getitem__(self, index):
        'Generate one batch of data'
        indexes = self.indexes[index*self.batch_size:(index+1)*self.batch_size]
        X = self.__data_generation(indexes)
        return X
            
    def __data_generation(self, indexes):
        'Generates data containing batch_size samples' 
        X = np.zeros((len(indexes),self.img_size,self.img_size,3),dtype='float32')
        df = self.df.iloc[indexes]
        for i,(index,row) in enumerate(df.iterrows()):
            img = cv2.imread(self.path+row.image)
            X[i,] = cv2.resize(img,(self.img_size,self.img_size)) [[/128]].0 - 1.0
        return X

JPG 파일 읽어오기

def parse_image(filename, image_size):  
	image = tf.io.read_file(filename)  
# decode_jpeg: Decode a JPEG-encoded Image to a Uint8 Tensor
	image = tf.image.decode_jpeg(image)  
# convert_image_dtype: Convert Image to Dtype, Scaling Its Values if Needed
	image = tf.image.convert_image_dtype(image, tf.float32)  
	if image_size is None:  
		image = tf.image.resize(image)  
	else:  
		image = tf.image.resize(image, image_size)  
	return image