Hi I’ve been going through the course but I’ve a question that I couldn’t find an answer yet. Suppose I have to train a model to classify multiple car models.
What I’d usually do would be to use Imagenet first and then train the model on the images of cars I want to classify. Maybe there is a big dataset of car images available online, but this big dataset doesn’t contain the car models I have in my dataset.
Could I use this “intermediate” dataset between the giant Imagenet and the really small own car images dataset in order to improve the model. It’d be sort of a two stage transfer learning but I don’t know if there’s already a name for it.
Is this feasible or there’s a problem lying there that I don’t see? Do you have some examples?