Using beautify/gender changing filters for data augmentation with facial recognition

This question really has two parts.
Firstly, has anyone used transformations of face images through different beautifying/Snapchat like filters to increase the number of photos and had good success. I specifically ask in relation to the kaggle competition trying to figure out if two people are relatives, and thought it could be clever to feed in photos of the person filtered to look like the opposite sex.

Secondly, is there a way to define a transform and add it into the standards get_transforms or somehow append a transform to the list. I couldn’t find what type of object a transform is in pyTorch easily so any direction here would be greatly appreciated.

Thanks in advance!!