What car is that?

WhatCar detects the make and model of a car from a photograph. It works using machine learning; an algorithm that takes bunch of photos of different objects and works out how to distinguish them. The model was trained using the Deep Learning techniques in fastai; it's easier than it sounds.

This was trained using cars commonly driven in Australia, so it may not work in the rest of the world. It recognises 400 of the most common cars in Australia, and gets it right more than 3 times in 4. This might not sound impressive, but some of the makes are very similar.

Thanks to fastai for teaching world class image recognition in a few hours, the Vehicle Make and Model Recognition Dataset (VMMRdb) for providing their data, Benjamin Ward for the background pattern, fuzzimo for the polaroid frames.