Portions of the Swift for TensorFlow ecosystem that might be missing for tutorials

[BradLarson]

With Swift for TensorFlow, and the Swift data science ecosystem in general, still being fairly early-stage, many elements of Python-based TensorFlow don't have equivalents on the Swift side. This looks like it might be an increasing issue with converting tutorials going forward, so it might be helpful to first list off libraries or capabilities that tutorials depend on but that don't exist yet.

I figured having an open issue like this might be a good way to track and discuss these, although we might be able to work this into a GitHub Project as well.

Here's the list that we've discussed so far, and feel free to add others:

• tensorflow_datasets -> would be great to leverage the existing work done here to make it easy to download / set up datasets
• image loading (tf.data, PIL / Pillow)
• simple image manipulation for augmentation (tensorflow.keras.preprocessing.image.ImageDataGenerator): flipping, shear, translation, cropping
• loading pretrained weights (TensorFlow Hub): Needs SavedModel loading, can currently pull from checkpoints via Raw TensorFlow operators