Sawatari Akane for PDXL and mixes.

The training data set was very small, but it seems to have picked up the prominent features well.

Note: when using `two-tone hair`, objects like hats will be displaced to show the hair roots.