I’m interested in building a “What would our baby look like?” app that takes as input two pictures and produces a baby picture with the features of the input pictures. I think this will be a good project for applying generative AI.
It can be fun, but who knows what will come out in a real life situation!
nice idea, good luck ,
i think an images of the parents parents will be helpful
Agreed. Genetics isn’t 100% predictable. But I think it would be an interesting exercise.
Yes. My thinking is to train on parents pictures and their resulting children. It would be hard to get the training data though.
I think mofahlih64 meant the model should train on images of both parents and grandparents to improve accuracy. Would be surprised if there was no prior research on whether this would help.
I suspect family tree sites like ancestry.com would have individual family member photos in their trees, but not sure if genealogy companies allow researchers access to images.
That’s kind of what I would think too. You probably could do that if you had enough data of a certain age like 18 years old for parents and children. Getting data from something like your grandparents might be difficult though since cameras weren’t digital then. Sometimes you only have a painting of grandparents. I guess your talking a 40 year window with two 18 year old parents and their 18 year old children.
firstly train the model on parents pictures and then resultant is children.