It seems like ML might make this possible, with some allowance for error
I was recently putting together a directory of people. People could include photos in their listing. People have…different ideas about what makes a good profile photo, but they almost always include their face somewhere. For my purposes, I needed all the photos at the same aspect ratio and at a certain minimum resolution.
So here’s what I’d like to do: create a recipe that detects a face, crops around that face with a certain margin and aspect ratio, and scales the crop to a certain resolution. Maybe create a report showing low-confidence crops.
Hello, and welcome!
I’m pretty sure what you want to do is possible with a custom JS plugin, and also using Apple APIs in the Vision framework (and along with Core Image).
We have documentation on the API here:
And a bunch of examples here:
This is a neat idea though, and maybe when things calm down a little bit over here I can take a stab at it.
Thanks. That’s above my level of expertise, and for my purposes, the juice wouldn’t be worth the squeeze.
I don’t know why I didn’t think of this - but there is the “Crop to Face” node under the Transform category. Maybe that would do what you’re after?
Huh. That does work. I’ll need to experiment to get good results.