Question

我正在尝试检测交通信号灯。我有3个标签的图像，所以我将训练数据和验证数据放入3个子文件夹中进行训练和验证。当我运行程序时，我在model.fit_generator（）时遇到错误。其他的事情都很好。

`

import os

from keras.layers import Dense, Dropout, Flatten, MaxPooling2D, 
Convolution2D
from keras.models import Sequential
from keras.preprocessing.image import ImageDataGenerator
import keras
from keras.optimizers import Adam

NUM_CHANNELS = 3
IMAGE_WIDTH = 224 # Original: 455
IMAGE_HEIGHT = 224 # Original: 256
NUM_CLASSES = 3
base_script_name = os.path.splitext(__file__)[0]
filepath=base_script_name + "-{epoch:02d}-val_acc-{val_acc:.2f}.hdf5"

def get_generator(directory, train):
if train:
datagen = ImageDataGenerator(
  rescale=1./255,
  shear_range=0.2,
  zoom_range=0.2,
  horizontal_flip=True)
else:
datagen = ImageDataGenerator(rescale=1./255)
return datagen.flow_from_directory(
directory=directory,
target_size=(IMAGE_WIDTH, IMAGE_HEIGHT),
batch_size=8,
class_mode='categorical')

model = Sequential([
Convolution2D(16, 3, 3, border_mode='same', subsample=(2, 2), input_shape=
(IMAGE_WIDTH, IMAGE_HEIGHT, NUM_CLASSES), activation='relu'),
MaxPooling2D(pool_size=(3, 3)),
Dropout(0.2),

Convolution2D(32, 3, 3, border_mode='same', activation='relu'),
MaxPooling2D(pool_size=(3, 3)),
Dropout(0.2),

Convolution2D(64, 3, 3, border_mode='same', activation='relu'),
MaxPooling2D(pool_size=(2, 2)),
Dropout(0.2),

Flatten(),
Dense(128, activation='tanh'),
Dropout(0.3),
Dense(NUM_CLASSES, activation='softmax'),
])
model.summary()

directory = '/home/nishat/traffic_light/nexar/try/data/'
train_generator = get_generator(directory+'/train', True)
validation_generator = get_generator(directory+'valid', False)

model.compile(optimizer=Adam(lr=0.0003), loss='categorical_crossentropy', 
metrics=['accuracy'])

# Callbacks
checkpoint = keras.callbacks.ModelCheckpoint(filepath, monitor='val_acc', 
verbose=1, save_best_only=False, mode='auto')
tensorboard = keras.callbacks.TensorBoard(log_dir='./tensorboar', 
histogram_freq=0, write_graph=True, write_images=True)
callbacks = [checkpoint, tensorboard]

model.fit_generator(
train_generator,
samples_per_epoch=train_generator.shape,
nb_epoch=200,
validation_data=validation_generator,
nb_val_samples=validation_generator.shape,
callbacks=callbacks,
)

model.evaluate_generator(validation_generator, 
val_samples=len(validation_generator.filenames))`

我收到如下错误：

Found 16791 images belonging to 3 classes.
Found 1868 images belonging to 3 classes.
Traceback (most recent call last):

  File "<ipython-input-1-303a170ea444>", line 1, in <module>
    runfile('/home/nishat/traffic_light/nexar/try/train.py', wdir='/home/nishat/traffic_light/nexar/try')

  File "/home/nishat/anaconda3/lib/python3.6/site-packages/spyder/utils/site/sitecustomize.py", line 866, in runfile
    execfile(filename, namespace)

  File "/home/nishat/anaconda3/lib/python3.6/site-packages/spyder/utils/site/sitecustomize.py", line 102, in execfile
    exec(compile(f.read(), filename, 'exec'), namespace)

  File "/home/nishat/traffic_light/nexar/try/train.py", line 69, in <module>
    samples_per_epoch=train_generator.shape,

AttributeError: 'DirectoryIterator' object has no attribute 'shape'

如果有人可以建议我需要更改我的代码，我将非常感激。感谢

Answer 1

两点：

1- train_generator和validation_generator没有'shape'属性

model.fit_generator( train_generator,
                     samples_per_epoch=train_generator,
                     nb_epoch=200,
                     validation_data=validation_generator,
                     nb_val_samples=validation_generator,
                     callbacks=callbacks)

2-最新的Keras版本'samples_per_epoch'重命名为'steps_per_epoch'，'nb_epoch'重命名为'epochs'

每个时期的步数可以按如下方式计算：

steps_per_epoch = int(math.ceil(float(train_ds_len) / batch_size))

在model.fit_generator（）上有错误

1 个答案: