Question

是否可以有两个fit_generator？

我创建了一个带有两个输入的模型，型号配置如下所示。

标签Y对X1和X2数据使用相同的标签。

将继续发生以下错误。

检查模型输入时出错：您传递给模型的Numpy数组列表不是模型预期的大小。预期看到2个数组，但得到以下1个数组的列表： [array（[[[[[0.75686276,0.75686276,0.75686276]， [0.75686276,0.75686276,0.75686276]， [0.75686276,0.75686276,0.75686276]， ... [0.65882355,0.65882355,0.65882355 ......

我的代码如下所示：

def generator_two_img(X1, X2, Y,batch_size):
    generator = ImageDataGenerator(rotation_range=15,
                                   width_shift_range=0.2,
                                   height_shift_range=0.2,
                                   shear_range=0.2,
                                   zoom_range=0.2,
                                   horizontal_flip=True,
                                   fill_mode='nearest')

    genX1 = generator.flow(X1, Y, batch_size=batch_size)
    genX2 = generator.flow(X2, Y, batch_size=batch_size)

    while True:
        X1 = genX1.__next__()
        X2 = genX2.__next__()
        yield [X1, X2], Y
  """
      .................................
  """
hist = model.fit_generator(generator_two_img(x_train, x_train_landmark, 
                y_train, batch_size),
                steps_per_epoch=len(x_train) // batch_size, epochs=nb_epoch,
                callbacks = callbacks,
                validation_data=(x_validation, y_validation),
                validation_steps=x_validation.shape[0] // batch_size, 
                `enter code here`verbose=1)

Answer 1

试试这个发电机：

def generator_two_img(X1, X2, y, batch_size):
    genX1 = gen.flow(X1, y,  batch_size=batch_size, seed=1)
    genX2 = gen.flow(X2, y, batch_size=batch_size, seed=1)
    while True:
        X1i = genX1.next()
        X2i = genX2.next()
        yield [X1i[0], X2i[0]], X1i[1]

在Thanh Nguyen评论之后编辑

3个输入的生成器：

def generator_two_img(X1, X2, X3, y, batch_size):
    genX1 = gen.flow(X1, y,  batch_size=batch_size, seed=1)
    genX2 = gen.flow(X2, y, batch_size=batch_size, seed=1)
    genX3 = gen.flow(X3, y, batch_size=batch_size, seed=1)
    while True:
        X1i = genX1.next()
        X2i = genX2.next()
        X3i = genX3.next()
        yield [X1i[0], X2i[0], X3i[0]], X1i[1]

Answer 2

我有一个TimeseriesGenerator的多个输入的实现，我已经对其进行了修改（不幸的是，我无法对其进行测试），以实现ImageDataGenerator的示例。我的方法是为keras.utils.Sequence的多个生成器构建一个包装器类，然后实现它的基本方法：__len__和__getitem__：

from keras.preprocessing.image import ImageDataGenerator
from keras.utils import Sequence


class MultipleInputGenerator(Sequence):
    """Wrapper of 2 ImageDataGenerator"""

    def __init__(self, X1, X2, Y, batch_size):
        # Keras generator
        self.generator = ImageDataGenerator(rotation_range=15, 
                                            width_shift_range=0.2,
                                            height_shift_range=0.2,
                                            shear_range=0.2,
                                            zoom_range=0.2,
                                            horizontal_flip=True, 
                                            fill_mode='nearest')

        # Real time multiple input data augmentation
        self.genX1 = self.generator.flow(X1, Y, batch_size=batch_size)
        self.genX2 = self.generator.flow(X2, Y, batch_size=batch_size)

    def __len__(self):
        """It is mandatory to implement it on Keras Sequence"""
        return self.genX1.__len__()

    def __getitem__(self, index):
        """Getting items from the 2 generators and packing them"""
        X1_batch, Y_batch = self.genX1.__getitem__(index)
        X2_batch, Y_batch = self.genX2.__getitem__(index)

        X_batch = [X1_batch, X2_batch]

        return X_batch, Y_batch

实例化生成器后，可以将其与model.fit_generator()一起使用。

Keras：如何使用fit_generator和多个输入

2 个答案: