I want to see how these new sneakers I'm about to buy will go with my (or similar to my) pants and shirt. We already have hundreds of thousands of models of different shapes, we already have pictures of merchandise in three planes (right on the official websites), why can't this be framed and combined?
Everything else starts on a computer in CAD, with great effect to making it accessible to people who aren't real big time enthusiasts, even a basic transparent PNG layering based approximation (Train an AI to take a picture of a piece of clothing and normalize it to a layer?) system would be really cool.
But now that I think about it, if you trained an AI like that you could probably make your own frontend site for Amazon that could take the pictures they give you and show a preview..
what would you want this front site over amazon to do in the preview?
I'm guessing it would be a real hassle to train though!