1
00:00:00,000 --> 00:00:04,362
[MUSIC]

2
00:00:04,362 --> 00:00:10,203
So I showed you some examples of
neural networks in computer vision and

3
00:00:10,203 --> 00:00:12,660
doing classification.

4
00:00:12,660 --> 00:00:15,112
Is there a labrador
retriever in this image?

5
00:00:15,112 --> 00:00:17,016
But they can do quite a bit more.

6
00:00:17,016 --> 00:00:20,760
So, for example, we can do image parsing.

7
00:00:20,760 --> 00:00:23,633
So in this example, for
every picture in the image,

8
00:00:23,633 --> 00:00:26,587
you're trying to classify it and
discover regions.

9
00:00:26,587 --> 00:00:30,027
So in the center top image,
you see a region of sky,

10
00:00:30,027 --> 00:00:33,030
another region of grass, and so on.

11
00:00:33,030 --> 00:00:36,880
And this kind of image description,
or is called scene understanding,

12
00:00:36,880 --> 00:00:41,730
is pretty cool, and you know networks
again, provided significant gains.

13
00:00:43,060 --> 00:00:46,440
But if we go back to
the beginning of the module

14
00:00:46,440 --> 00:00:50,080
when we discussed a new way to shop for
shoes or dresses,

15
00:00:50,080 --> 00:00:54,110
the thing that we're really trying to
do there is retrieve similar images.

16
00:00:54,110 --> 00:00:58,130
So, for example, if I give you
the input of this boring black shoe,

17
00:00:58,130 --> 00:01:00,170
what neural network with output.

18
00:01:00,170 --> 00:01:05,007
Deep neural network is
the shoes that are shown here.

19
00:01:05,007 --> 00:01:06,900
A bunch of boring black shoes.

20
00:01:06,900 --> 00:01:09,800
Now, if I do this a little
bit more stylish boot,

21
00:01:09,800 --> 00:01:14,710
you'll see that it gives you a variety
of interesting boots like that.

22
00:01:14,710 --> 00:01:19,770
Similarly, for heels, for
brown shoes, for sneakers and so on.

23
00:01:19,770 --> 00:01:24,237
So this is the core of the concept that
we're using in the demo that we showed at

24
00:01:24,237 --> 00:01:25,976
the beginning of the module.

25
00:01:25,976 --> 00:01:29,979
[MUSIC]