https://lucid.app/documents/embeddedchart/9d72fa85-b820-4ce7-abdf-b7628f31e02c
- upload photo (orient via LoadImage)
- detect face(s) - set relative unit
- get foreground/background (grabcut)
- set roi for skin, lips, hair, eyes
- get kmeans palette / expand via inrange
- refine mask via grabcut
- colorize (!greenscreen)
- composite color mask (?)
- composite depth mask (PIXI)
- bar chart (normal) and hsl color wheel (d3)
Pre-visualize segmentation and modification for benchmark and/or a more robust application in 3d?