I've wanted to build this side-project for years - but training a model on every car was prohibitively expensive and complicated to do. It became trivial with the launch of GPT-4 Vision API. Some iteration on the prompt was required until it became pretty dang accurate. App is built 100% in SwiftUI. Would love your feedback on usefulness and price (its free up to 3 identifications).
I downloaded to try it out, is impressively accurate. Even told me the model backhoe I was looking at! I can't believe there wasn't already a successful app in this niche.<p>I can't shake that this would have felt like magic 5 years ago, but now it's so casual and not even that impressive. My mind is like "cool, now what?". I guess cars aren't that complicated.