Convolutional neural networks filter training

Hi Everyone,

I have two major points that I'm trying to understand in CNN. The first one: Assume that we have 500 testing images. Are we going to feed them to the filters/kernels all at once for convolution. Or its done input test image one after the other? For the second question: Do we have a stack of filters/kernels for each input test image or it’s a single stack for all input images. Lets say we are using 12 filters, are they 12 for each input test image or for the whole?

Thanks a lot in advance.