I've generated a few that
you can use to test, they're at the URL in the previous lesson
with the training set, test set, and validation
set are all link to. Download the RPS validation zip from the link shown and in it, you'll find 33 images of different hand models ones
that weren't using the training and validation sets and they're in different poses. Let's now upload one of these to the workbook and we'll
try and get what happens. I'm going to pick a paper pose, and the classification
comes back as 1-0-0. When using the image generator, the classes come
from directories and thus were sorted in
alphabetical order. So the first value is for paper and then rock
and then scissors. So we can see that
this is correct. So if I try another
one a scissors, we'll see that
the third neuron lights up which is what I would expect. Lets now try a rock, and this one is a larger picture hence the high resin filename, and I want you to see
how that would perform. It's slower to upload and
it needs to re-scale, but we can see that it gets classified
correctly as a rock. I can also upload all the files at once and see
classifications for them all. So let's try that. I'm speeding up the video to get
past all the uploads. So let's take a look at the classifications
that I get back. Remember that the neurons are in the order paper then rock then scissors because
it's alphabetical. So we can see that the first one is
right. The third one is correct and the same with
the fourth and the fifth. So let's keep looking
through the list. They mostly seem to be correct. The rest of the data looks like it was
classified correctly. So that was a look at
multiclass classification. Hopefully, you found
it interesting.