What makes it a good dataset?

[cherishh]

Hi, very inspiring work! Now I'm trying to do the same on bumble, just to learn more about model training. I've wrote a script that when i swipe right/left, it will collecting relative image urls & put them into one of two json files(likes/dislikes). Works just fine. My problem is i don't have any experience in AI or model training. So could you please tell me, what makes it a good dataset? Now I just collect all the images i encounter. But after this, should i just extract the faces? should i make all images the same size? some of the images are kind of higher quality, some of them are blurry, what should i do with this situation? Thx a lot!!