Is there a way to train with one class of images only?

I want to train, for example, with just cats and later be able to predict if the images are cats or not. Is this possible and how? Thanks.

I have the same question!

Can i use a pre trained resnet (or another model) and train it to recognize one class vs all? like its either a cat or not.


Yes you can :slight_smile: the best way to do it is to think about what specially you want as “not cat”? Will users be inputting animals to your app? Then get a few pictures of a variety of animals and label them ‘not cat’.

Well i used the cat vs not cat as an example. In reality im trying to make an app that checks if there is some inappropriate(lets say nudes) content on the screen of a tablet/phone. So my class would be either:

1.- inappropriate
2.- anything else

So thats my problem, how do I label the 2nd class since resnet is already trained to recognize tons of things.