1
00:00:00,000 --> 00:00:03,030
I've generated a few that
you can use to test,

2
00:00:03,030 --> 00:00:04,530
they're at the URL in

3
00:00:04,530 --> 00:00:06,170
the previous lesson
with the training set,

4
00:00:06,170 --> 00:00:08,775
test set, and validation
set are all link to.

5
00:00:08,775 --> 00:00:11,100
Download the RPS validation zip

6
00:00:11,100 --> 00:00:12,570
from the link shown and in it,

7
00:00:12,570 --> 00:00:14,640
you'll find 33 images of

8
00:00:14,640 --> 00:00:17,010
different hand models ones
that weren't using

9
00:00:17,010 --> 00:00:18,510
the training and validation sets

10
00:00:18,510 --> 00:00:20,160
and they're in different poses.

11
00:00:20,160 --> 00:00:22,560
Let's now upload one of these to

12
00:00:22,560 --> 00:00:24,825
the workbook and we'll
try and get what happens.

13
00:00:24,825 --> 00:00:27,550
I'm going to pick a paper pose,

14
00:00:28,880 --> 00:00:32,910
and the classification
comes back as 1-0-0.

15
00:00:32,910 --> 00:00:35,220
When using the image generator,

16
00:00:35,220 --> 00:00:37,050
the classes come
from directories and

17
00:00:37,050 --> 00:00:39,450
thus were sorted in
alphabetical order.

18
00:00:39,450 --> 00:00:40,850
So the first value is for

19
00:00:40,850 --> 00:00:43,250
paper and then rock
and then scissors.

20
00:00:43,250 --> 00:00:45,575
So we can see that
this is correct.

21
00:00:45,575 --> 00:00:48,890
So if I try another
one a scissors,

22
00:00:48,890 --> 00:00:51,440
we'll see that
the third neuron lights

23
00:00:51,440 --> 00:00:53,330
up which is what I would expect.

24
00:00:53,330 --> 00:00:54,890
Lets now try a rock,

25
00:00:54,890 --> 00:00:56,810
and this one is a larger picture

26
00:00:56,810 --> 00:00:58,639
hence the high resin filename,

27
00:00:58,639 --> 00:01:00,710
and I want you to see
how that would perform.

28
00:01:00,710 --> 00:01:03,900
It's slower to upload and
it needs to re-scale,

29
00:01:03,900 --> 00:01:05,870
but we can see that

30
00:01:05,870 --> 00:01:08,225
it gets classified
correctly as a rock.

31
00:01:08,225 --> 00:01:10,640
I can also upload all the files

32
00:01:10,640 --> 00:01:12,830
at once and see
classifications for them all.

33
00:01:12,830 --> 00:01:15,020
So let's try that. I'm speeding

34
00:01:15,020 --> 00:01:17,030
up the video to get
past all the uploads.

35
00:01:17,030 --> 00:01:18,110
So let's take a look at

36
00:01:18,110 --> 00:01:19,865
the classifications
that I get back.

37
00:01:19,865 --> 00:01:21,650
Remember that the neurons are in

38
00:01:21,650 --> 00:01:23,250
the order paper then rock

39
00:01:23,250 --> 00:01:25,280
then scissors because
it's alphabetical.

40
00:01:25,280 --> 00:01:26,630
So we can see that

41
00:01:26,630 --> 00:01:28,030
the first one is
right.

42
00:01:28,030 --> 00:01:30,010
The third one is correct

43
00:01:30,010 --> 00:01:32,355
and the same with
the fourth and the fifth.

44
00:01:32,355 --> 00:01:34,300
So let's keep looking
through the list.

45
00:01:34,300 --> 00:01:36,265
They mostly seem to be correct.

46
00:01:43,725 --> 00:01:45,650
The rest of the data

47
00:01:45,650 --> 00:01:47,605
looks like it was
classified correctly.

48
00:01:47,605 --> 00:01:50,300
So that was a look at
multiclass classification.

49
00:01:50,300 --> 00:01:52,620
Hopefully, you found
it interesting.