def read(self, img_rows=IMAGE_SIZE1, img_cols=IMAGE_SIZE2):
train_datagen = ImageDataGenerator(rescale=1. / 255)
valid_datagen = ImageDataGenerator(rescale=1. / 255)
test_datagen = ImageDataGenerator(rescale=1. / 255)
train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size=(img_rows, img_cols),
batch_size=batch_size,
class_mode='binary')
validation_generator = valid_datagen.flow_from_directory(
validation_data_dir,
target_size=(img_rows, img_cols),
batch_size=batch_size,
class_mode='binary')
test_generator = test_datagen.flow_from_directory(
test_data_dir,
target_size=(img_rows, img_cols),
batch_size=batch_size,
class_mode='binary')
self.train = train_generator
self.valid = validation_generator
self.test = test_generator
def __init__(self,
train_dir,
validate_dir,
batch_size=100,
data_augmentation=True):
if data_augmentation:
print("Enable data augmentation")
train_datagen = ImageDataGenerator(
#rescale=1./255,
preprocessing_function=lambda x: x / 255 - 0.5,
shear_range=0.2,
zoom_range=0.2,
width_shift_range=0.3,
height_shift_range=0.3,
horizontal_flip=True,
fill_mode='nearest')
else:
print("Disable data augmentation")
train_datagen = ImageDataGenerator(
preprocessing_function=lambda x: x / 255 - 0.5)
validation_datagen = ImageDataGenerator(
preprocessing_function=lambda x: x / 255 - 0.5)
train_generator_flow = train_datagen.flow_from_directory(
train_dir,
target_size=(299, 299),
batch_size=batch_size,
class_mode="input")
# this is a similar generator, for validation data
validation_generator_flow = validation_datagen.flow_from_directory(
validate_dir,
target_size=(299, 299),
batch_size=batch_size,
class_mode="input")
self.train_generator_flow = train_generator_flow
self.validation_generator_flow = validation_generator_flow
def generator_for_input(data_dir, batch_size, img_rows, img_cols, seed=0):
print("generator_for_mulinputs params :", data_dir, batch_size, img_rows,
img_cols)
datagen = ImageDataGenerator(rescale=1. / 255, horizontal_flip=True)
print("all propertys of datagen", datagen)
# 打印对象的所有属性值的方法
# print("\n".join(['%s:%s' % item for item in train_datagen.__dict__.items()]))
gen = datagen.flow_from_directory(
data_dir,
target_size=(img_rows, img_cols),
batch_size=batch_size,
class_mode='categorical', # 多分类; 'binary')
seed=seed) # seed必须一致，生成的数据顺序前后才是一样的。
# print("gen.filenames", gen.filenames) # 按顺序输出文件的名字
print("gen.class_indices", gen.class_indices) # 输出对应的标签文件夹
while True:
inputx, inputy = next(gen)
yield inputx, inputy
test_data_dir = 'data/miniimagenet/val' #
img_rows = 224 # 227
img_cols = 224 # 227
epochs = 3
# 批量大小
batch_size = 4
# 训练样本总数
nb_train_samples = 3120 #3*1040
#all num of val samples
nb_validation_samples = 780 #3*260
##################数据导入及预处理###########################
train_datagen = ImageDataGenerator(rescale=1. / 255, horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1. / 255)
train_generator = train_datagen.flow_from_directory(
train_data_dir,
target_size=(img_rows, img_cols),
batch_size=batch_size,
class_mode='categorical') #多分类; 'binary')
print("train_generator.filenames", train_generator.filenames) # 按顺序输出文件的名字
print("train_generator.class_indices",
train_generator.class_indices) #输出对应的标签文件夹
validation_generator = test_datagen.flow_from_directory(
test_data_dir,
target_size=(img_rows, img_cols),
batch_size=batch_size,
class_mode='categorical') #多分类; 'binary')
print("validation_generator.filenames",
validation_generator.filenames) # 按顺序输出文件的名字
print("validation_generator.class_indices",
validation_generator.class_indices) #输出对应的标签文件夹
##################网络搭建###########################
classifier.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Part 2 - Fitting the CNN to the images
batch_size = 32
train_datagen = ImageDataGenerator(rescale=1. / 255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1. / 255)
training_set = train_datagen.flow_from_directory(train_set_path,
target_size=input_size,
batch_size=batch_size,
class_mode='binary')
test_set = test_datagen.flow_from_directory(test_set_path,
target_size=input_size,
batch_size=batch_size,
class_mode='binary')
# Create a loss history
history = LossHistory()
# train model
classifier.fit_generator(training_set,
steps_per_epoch=8000 / batch_size,
epochs=90,
validation_data=test_set,
classifier.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['accuracy'])
#ağı öğretmek için her epochste datasetten kaçar tane resim alacak
batch_size = 10
train_datagen = ImageDataGenerator(rescale = 1./255,
shear_range = 0.2,
zoom_range = 0.2,
horizontal_flip = True)
validation_datagen = ImageDataGenerator(rescale=1./255,validation_split=0.10)
test_datagen = ImageDataGenerator(rescale=1./ 255)
training_set = train_datagen.flow_from_directory(training_set_path,
target_size=input_size,
batch_size=batch_size,
subset="training",
class_mode='categorical')
validation_set = validation_datagen.flow_from_directory(validation_set_path,
target_size=input_size,
batch_size=batch_size,
subset="validation",
class_mode='categorical')
test_set = test_datagen.flow_from_directory(test_set_path,
target_size=input_size,
color_mode="rgb",
shuffle = False,
batch_size=1,
class_mode='categorical')
# Compiling the CNN
classifier.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
# Part 2 - Fitting the CNN to the images
batch_size = 32
train_datagen = ImageDataGenerator(rescale=1. / 255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1. / 255)
training_set = train_datagen.flow_from_directory('dataset/training_set',
target_size=input_size,
batch_size=batch_size,
class_mode='binary')
test_set = test_datagen.flow_from_directory('dataset/test_set',
target_size=input_size,
batch_size=batch_size,
class_mode='binary')
# Create a loss history
class LossHistory(Callback):
def __init__(self):
super().__init__()
self.epoch_id = 0
self.losses = ''
def add_data(model, train_images, train_labels, test_images, test_labels, train_ratio):
from tensorflow.contrib.keras.api.keras.preprocessing.image import ImageDataGenerator
image_size = int(config['DEFAULT']['IMAGE_SIZE'])
datagen = ImageDataGenerator()
generator = datagen.flow_from_directory(
utils.CATEGORIES_LOCATION,
color_mode='grayscale',
target_size=(image_size, image_size),
batch_size=1,
class_mode='binary')
number_of_images = generator.samples
number_of_categories = generator.num_classes
number_processed = 0
images = []
labels = []
# is there any data?
if number_of_images == 0:
return None
# stores how many images of each category are present
number_per_category = {c: 0.0 for c in range(number_of_categories)}
while number_processed < number_of_images:
item = next(generator)
image = numpy.array(item[0], dtype=numpy.uint8).reshape(1, image_size, image_size, 1)
if model == "regression":
image = ((255 - image) / 255.0)
elif model == "CNN":
image = (((255 - image) / 255.0) - 0.5)
image = numpy.reshape(image, [1, -1])
label = int(item[1][0])
number_per_category[label] += 1.0
labels.append(label)
images.append(numpy.reshape(image, image_size * image_size))
number_processed += 1
# Code with test set
# stores how many images of each category are in the training set
#number_per_category_in_training = {c: 0.0 for c in range(NUM_LABELS)}
for i, x in enumerate(images):
# Code without test set
train_images.append(x)
train_labels.append(labels[i])
# Code with test set
#if number_per_category_in_training[category] < number_per_category[category] * train_ratio:
# number_per_category_in_training[category] += 1.0
# train_images.append(x)
# train_labels.append(labels[i])
#else:
# test_images.append(x)
# test_labels.append(labels[i])
train_images = numpy.array(train_images)
# Code with test set
#test_images = numpy.array(test_images)
# transform labels into one-hot vectors
one_hot_encoding = numpy.zeros((len(train_images), number_of_categories))
one_hot_encoding[numpy.arange(len(train_images)), train_labels] = 1
train_labels = numpy.reshape(one_hot_encoding, [-1, number_of_categories])
# Code with test set
#one_hot_encoding = numpy.zeros((len(test_images), number_of_categories))
#one_hot_encoding[numpy.arange(len(test_images)), test_labels] = 1
#test_labels = numpy.reshape(one_hot_encoding, [-1, number_of_categories])
# Code without test set
return train_images, train_labels, None, None
classifier.add(Dense(units=1, activation='sigmoid'))
# Compiling the CNN
classifier.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])
# Part 2 - Fitting the CNN to the images
batch_size = 32
train_datagen = ImageDataGenerator(rescale=1. / 255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1. / 255)
training_set = train_datagen.flow_from_directory('G:\\ANALYTICS_WORLD_R_SAS\\python_world\\deep learning\\Convolutional_Neural_Networks\\dataset\\training_set',
target_size=input_size,
batch_size=batch_size,
class_mode='binary')
test_set = test_datagen.flow_from_directory('G:\\ANALYTICS_WORLD_R_SAS\\python_world\\deep learning\\Convolutional_Neural_Networks\\dataset\\test_set',
target_size=input_size,
batch_size=batch_size,
class_mode='binary')
# Create a loss history
history = LossHistory()
classifier.fit_generator(training_set,
steps_per_epoch=8000/batch_size,
epochs=90,
validation_data=test_set,
validation_steps=2000/batch_size,
script_dir = os.path.dirname(".")
training_set_path = os.path.join(script_dir, '/content/drive/My Drive/train/ProjTrain/')
test_set_path = os.path.join(script_dir, '/content/drive/My Drive/train/ProjTrain/')
batch_size = 32
input_size = (512,512)
train_datagen = ImageDataGenerator(rescale=1. / 255,
shear_range=0.2,
zoom_range=0.2,
horizontal_flip=True)
test_datagen = ImageDataGenerator(rescale=1. / 255, validation_split=0.33)
training_set = train_datagen.flow_from_directory(training_set_path,
target_size=input_size,
batch_size=batch_size,
subset="training",
class_mode='categorical')
test_set = test_datagen.flow_from_directory(test_set_path,
target_size=input_size,
batch_size=batch_size,
subset="validation",
class_mode='categorical')
#########################
### Build the model