Hi, sorry if this is a novice question, I'm a junior who is new to machine learning but is super interested. For a school project I am trying to recommend clothing images, I get how to tell if a image is a shirt or shoe, take a dataset of images with each image labeled shirt or shoe, then train a model and use it to classify a new image. I don't get how to do this when there's no specific label, like if you want to show more shoes that are similar in color. I am thinking to use nearest neighbor, but how would this work and what metric? Can neural networks do this?
[link][8 comments]