How do I completely ignore labels when using Keras image_dataset_from_directory?

Question

When trying to load my image dataset for an unsupervised problem, without labels, the image_dataset_from_directory function from Keras assumes they belong to a class. The resulting dataset has shape:

(batch_size, None, image_height, image_width, n_channels)

The code:

data_dir = pathlib.Path(path_to_image)

train_dataset = (keras.utils.image_dataset_from_directory
                  (
                    data_dir,
                    labels=None,
                    validation_split=validation_split,
                    subset="training",
                    shuffle=False,
                    image_size=(image_size, image_size)
                  )
            
            .map(preprocess_image, num_parallel_calls=tf.data.AUTOTUNE)
            .cache()
            .batch(batch_size, drop_remainder=True)
            .prefetch(buffer_size=tf.data.AUTOTUNE))

I believe the dimension with None should be 2 in the case of a dataset with both inputs and target pairs. However, I only wish to load my images from a folder, without labels.