I would like to build an app that recommends the most similar object to an image of a given object. The user would input an image from their camera roll, and the app would output similar images sourced from the internet. What would be the best tech stack to build this product?