1
00:00:00,000 --> 00:00:04,124
[MUSIC]

2
00:00:04,124 --> 00:00:08,823
So we saw that deep learning had
a tremendous part in the ImageNet

3
00:00:08,823 --> 00:00:10,450
competition.

4
00:00:10,450 --> 00:00:15,030
Which allowed them to take 1.5 minute
image string deeply on your network and

5
00:00:15,030 --> 00:00:18,980
get amazing performance to predict one
of a thousand different categories.

6
00:00:18,980 --> 00:00:21,910
So let's go ahead and
show you a little demo of what kind

7
00:00:21,910 --> 00:00:25,420
of categories we're talking about and
how cool the predictions were.

8
00:00:25,420 --> 00:00:28,100
So here's an example.

9
00:00:28,100 --> 00:00:32,800
It was the AlexNet frame
on that ImageNet data set,

10
00:00:32,800 --> 00:00:38,420
which we then employed as a service
that can be queried from this website.

11
00:00:38,420 --> 00:00:41,570
And so every time I click on an image
it gets sent to that service

12
00:00:41,570 --> 00:00:45,730
which actually runs on a GPU, so it's
fast and it comes back for prediction.

13
00:00:45,730 --> 00:00:48,020
So if I click on this
particular image here,

14
00:00:48,020 --> 00:00:51,310
it gets sent to a service that
actually hosts in on Amazon AWS.

15
00:00:51,310 --> 00:00:52,560
It comes back for prediction here.

16
00:00:52,560 --> 00:00:55,740
It's hidden, but when I click on it,
it tells me what prediction is.

17
00:00:55,740 --> 00:00:59,580
So if I show you this image, it might
be unclear what that image is, but

18
00:00:59,580 --> 00:01:03,250
if I click on it, it says parking meter,
it turns out to be the right label.

19
00:01:04,310 --> 00:01:08,820
The second best prediction was padlock,
Which you can see kind of a padlock.

20
00:01:08,820 --> 00:01:10,340
The parking meter, you got it right.

21
00:01:10,340 --> 00:01:12,850
So that's really quite cool.

22
00:01:12,850 --> 00:01:14,690
Let me show you another example.

23
00:01:14,690 --> 00:01:15,850
For example, this one.

24
00:01:15,850 --> 00:01:20,110
It get shipped off to the service on
Amazon WS, comes back for prediction.

25
00:01:20,110 --> 00:01:23,761
And here my prediction, screen, monitor,
or [INAUDIBLE] it says it's a monitor, but

26
00:01:23,761 --> 00:01:27,880
I don't know what the difference between a
screen and a monitor is, but that's okay.

27
00:01:27,880 --> 00:01:32,170
So there's various images here,
I'm just gonna click on a few.

28
00:01:32,170 --> 00:01:37,170
So for example, if I click on this
one over here, it gets sent up.

29
00:01:37,170 --> 00:01:40,230
It's really sure,
it really thinks it's a spoonbill,

30
00:01:40,230 --> 00:01:44,087
it turns out to be a spoonbill,
which is great.

31
00:01:44,087 --> 00:01:49,700
Lastly I'm gonna click on
this one over here and

32
00:01:49,700 --> 00:01:54,460
that image gets sent to that service
uses a deep learning for GPU and

33
00:01:55,460 --> 00:02:00,060
it says it's a beer bottle or pop bottle,
the true label is beer bottle.

34
00:02:00,060 --> 00:02:03,140
And now this all image is in
the original ImageNet data set,

35
00:02:03,140 --> 00:02:06,310
I'm sure an image that was not in
the original ImageNet data set,

36
00:02:06,310 --> 00:02:09,198
I click on this one here,
it gets sent to that service.

37
00:02:09,198 --> 00:02:14,000
On the AWS that we're hosting there.

38
00:02:14,000 --> 00:02:19,970
Comes back for prediction, it says,
Labrador retriever, this is my dog.

39
00:02:19,970 --> 00:02:23,059
This is the lab in
>> [LAUGH] And

40
00:02:23,059 --> 00:02:26,480
this is the dog over here, in Dato.

41
00:02:26,480 --> 00:02:31,690
And so, as you can see, even for images
that were not in the original data set,

42
00:02:31,690 --> 00:02:34,470
you can still get pretty
interesting predictions.

43
00:02:34,470 --> 00:02:39,120
Now in your capstone,
you want to build a service now not for

44
00:02:39,120 --> 00:02:42,910
predicting images here but for
recommendations with deep learning, for

45
00:02:42,910 --> 00:02:46,140
product images and text and
host it as a service.

46
00:02:46,140 --> 00:02:49,582
And you'll be able to get a website like
this, that anybody can play with, use, and

47
00:02:49,582 --> 00:02:52,849
really see the power of the machine
learning that you've been learning about.

48
00:02:52,849 --> 00:02:57,589
[MUSIC]