Keras: Multiple Inputs for fit_generator using flow_from_directory

user_6249001 0
Viewed 2309

Is it possible to have two flow_from_directory in a fit_generator?

For example

train_generator_1 = train_datagen.flow_from_directory(
    train_data_dir_1,
    target_size=(img_height, img_width),
    batch_size=batch_size,
    class_mode='binary')

train_generator_2 = train_datagen.flow_from_directory(
    train_data_dir_2,
    target_size=(img_height, img_width),
    batch_size=batch_size,
    class_mode='binary')

custom_vgg_model.fit_generator(
    [train_generator_1, train_generator_2],
    steps_per_epoch=nb_train_samples,
    epochs=epochs,
    validation_data=validation_generator,
    validation_steps=nb_validation_samples,
    callbacks=[checkpointer, csv_logger])
keras tensorflow python
1 Answers

You can try making a generator that contains both:

In python 3: 

def doubleGenerator(generator1,generator2):

    while True:
        for (x1,y1),(x2,y2) in zip(generator1,generator2):
            yield ([x1,x2],y1)
user_20972400
Related