Let me put my thought of CNN training:

Lets say you want your model to recogonize one picture to tell us if its a cat. When we train the model, we need to input 100 cat pictures for the CNN for traning. When the model is trained, we input ONE picture, the model will tell us if its a cat. Am I right?

If I am right, I have a doult: if the pictures a big, lets say 512512 for each picture. So we really need to input 100 512512 pictures to the model?

Or I did not get the point of CNN?

Thanks.

I read the theory books. but not sure if I understood it.