Keras: How to use fit_generator with multiple inputs

Question

Is it possible to have two fit_generator?

I\'m creating a model with two inputs, The model configuration is shown below.

Label Y uses the same label

Answer 1

I have an implementation for multiple inputs for TimeseriesGenerator that I have adapted it (I have not been able to test it unfortunately) to meet this example with ImageDataGenerator. My approach was to build a wrapper class for the multiple generators from keras.utils.Sequence and then implement the base methods of it: __len__ and __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

You can use this generator with model.fit_generator() once the generator has been instanced.