Pretrained neural networks to recognize certain things

How to approach the problem of is it or is not classification? For example, I want to recognize if a face has makeup, but to do this I first need to recognize if it a face. Since a very small amount of pictures are face I can’t possibly feed the no_face category with everything else. Also is it better to recognize if it a face first and then second neural network to recognize if it is with makeup?

Are there pretrained models I can download for the most common things like faces, animals and so on?