1
00:00:00,000 --> 00:00:04,157
[MUSIC]

2
00:00:04,157 --> 00:00:07,591
Let's dig in and talk a little bit
about the capstone application

3
00:00:07,591 --> 00:00:09,730
that we're gonna build together.

4
00:00:09,730 --> 00:00:14,670
So we're gonna build a recommender
system for products that's intelligent,

5
00:00:14,670 --> 00:00:18,550
that combines text, images,
sentiment analysis, and deep learning.

6
00:00:18,550 --> 00:00:21,200
Now, before talking about the capstone
project which is really going

7
00:00:21,200 --> 00:00:21,770
to be exciting.

8
00:00:21,770 --> 00:00:24,870
Let me give you a quick demo of
something that you could build,

9
00:00:24,870 --> 00:00:28,260
a recommender system that you could build,
that combines text and image data and

10
00:00:28,260 --> 00:00:30,340
uses deep learning at its core.

11
00:00:30,340 --> 00:00:34,520
Now let me show you an example of an
intelligent application for recommending

12
00:00:34,520 --> 00:00:40,330
products that combines image and text
data and uses deep learning at its core.

13
00:00:40,330 --> 00:00:43,550
So let me start off with a story.

14
00:00:43,550 --> 00:00:48,480
So my sister's a fashion designer, and
her birthday was coming up this month, and

15
00:00:48,480 --> 00:00:50,550
I wanted to get her a gift
that would impress her.

16
00:00:50,550 --> 00:00:52,730
I want to get her a dress.

17
00:00:52,730 --> 00:00:56,180
But it's really hard for
me to choose a good dress.

18
00:00:56,180 --> 00:00:57,680
I don't know much about dresses.

19
00:00:57,680 --> 00:01:02,480
So if I do just standard keyword
search and I search for dresses.

20
00:01:02,480 --> 00:01:06,760
I am going to end up with something that
looks like this that I am showing you

21
00:01:06,760 --> 00:01:07,877
on the right here.

22
00:01:07,877 --> 00:01:12,641
And you will see there's a wide range
of dresses and it is very hard for

23
00:01:12,641 --> 00:01:16,584
me to describe the kind of dress
that my sister likes and so

24
00:01:16,584 --> 00:01:20,281
I have seen her for
example wear some floral dresses so

25
00:01:20,281 --> 00:01:25,375
I could of course search that combines
the keywords dress and floral, and

26
00:01:25,375 --> 00:01:31,380
looks through the description of the dress
and tries to find some floral dresses.

27
00:01:31,380 --> 00:01:33,360
But even within floral dresses,

28
00:01:33,360 --> 00:01:37,300
you see that there is a wide
range of floral dresses up there.

29
00:01:37,300 --> 00:01:41,830
And each one of them has
a pretty different kind of style.

30
00:01:41,830 --> 00:01:47,460
So the question is, how can I describe
what is a good style for my sister?

31
00:01:47,460 --> 00:01:50,840
Now, for example,
take this dress over here.

32
00:01:50,840 --> 00:01:53,890
It looks a little bit
like what she might like.

33
00:01:53,890 --> 00:01:55,400
Those warm sunset colors and rich.

34
00:01:55,400 --> 00:01:58,510
So there's some text analyses
that we can do that shows you

35
00:01:58,510 --> 00:02:01,720
a little bit about what describes this
dress but really visually looks good.

36
00:02:01,720 --> 00:02:05,730
So the question is can we use visual
cues to find other dresses that look

37
00:02:05,730 --> 00:02:07,090
kind of like this one?

38
00:02:07,090 --> 00:02:09,530
And you do things like this
in your capstone project.

39
00:02:09,530 --> 00:02:10,930
When I click on this dress,

40
00:02:10,930 --> 00:02:14,550
we're gonna find visually similar dresses
using a technique called deep learning.

41
00:02:14,550 --> 00:02:18,030
Then you're gonna learn about
already in this first course and

42
00:02:18,030 --> 00:02:20,090
apply it in practice.

43
00:02:20,090 --> 00:02:23,130
So when I click on this dress,
what I'm gonna see

44
00:02:23,130 --> 00:02:26,560
on the right is a bunch of dresses
that are visually similar to this one.

45
00:02:27,570 --> 00:02:32,080
And you will see, for example,
the dresses tend to be floral,

46
00:02:32,080 --> 00:02:34,530
they tend to have the similar hues,

47
00:02:34,530 --> 00:02:39,370
similar colors, and even then, it's hard
to describe what my sister might like.

48
00:02:39,370 --> 00:02:41,360
But I can look at this dress
right in the middle over here,

49
00:02:41,360 --> 00:02:45,140
and say wow, this has Interesting pattern,
different colors.

50
00:02:45,140 --> 00:02:46,700
Maybe this is what my sister likes.

51
00:02:46,700 --> 00:02:50,920
And if I click on it, I'm gonna find
visually similar dresses to this one, so

52
00:02:50,920 --> 00:02:53,050
dresses that have more
interesting patterns and

53
00:02:53,050 --> 00:02:55,360
multiple colors associated with them.

54
00:02:55,360 --> 00:02:58,420
So you'll see as I scroll,
there's variants of these.

55
00:02:58,420 --> 00:03:01,320
And then maybe my eye
catches this one over here.

56
00:03:01,320 --> 00:03:04,850
Where it looks like something
that my sister might like and

57
00:03:04,850 --> 00:03:08,680
I might call her up and say,
hey I think I found a great dress for you.

58
00:03:08,680 --> 00:03:10,740
This one.

59
00:03:12,447 --> 00:03:17,840
Now let's say that I call my sister and
she looks kinda like this model.

60
00:03:17,840 --> 00:03:20,970
I call her up and she says, you know.

61
00:03:20,970 --> 00:03:24,370
This looks okay, but
I am going to a cocktail party and

62
00:03:24,370 --> 00:03:26,800
why don't you get me a cocktail dress.

63
00:03:26,800 --> 00:03:30,662
So, I am trying to think about,
I start over.

64
00:03:30,662 --> 00:03:35,390
And I'm going to try to
find her a cocktail dress.

65
00:03:38,760 --> 00:03:41,570
Cocktail dresses are ones
that you wear for

66
00:03:41,570 --> 00:03:45,110
more formal parties that
look kind of like this.

67
00:03:45,110 --> 00:03:47,290
And there's all sorts of colors and
things.

68
00:03:47,290 --> 00:03:49,510
But maybe she's interested
in a black cocktail dress.

69
00:03:49,510 --> 00:03:56,020
So just play some keyword search like
black cocktail dress, you'll get things

70
00:03:56,020 --> 00:03:59,380
like this, but it's hard to describe
what my sister might be interested in.

71
00:03:59,380 --> 00:04:00,850
So I might ask her.

72
00:04:00,850 --> 00:04:01,670
What you're interested in.

73
00:04:01,670 --> 00:04:05,410
She says you know what I'm interested in
is a a cocktail dress that's interesting,

74
00:04:05,410 --> 00:04:06,350
with a touch of color.

75
00:04:07,940 --> 00:04:08,730
She's a fashion designer.

76
00:04:09,780 --> 00:04:11,060
What do I do?

77
00:04:11,060 --> 00:04:12,690
How do I describe that?

78
00:04:12,690 --> 00:04:14,100
So I look for these and say well,

79
00:04:14,100 --> 00:04:16,430
this first few on the left
is kind of interesting.

80
00:04:16,430 --> 00:04:20,870
If you look at the keywords associated
with it, you see the keyword jazzy.

81
00:04:20,870 --> 00:04:23,440
So maybe she's interested
in a jazzy dress.

82
00:04:23,440 --> 00:04:25,220
So what do jazzy dresses look like?

83
00:04:25,220 --> 00:04:28,420
So when I click on this keyword,
that I discover from the text,

84
00:04:28,420 --> 00:04:31,210
you see other jazzy dresses and
they're kind of interesting but

85
00:04:31,210 --> 00:04:33,730
they're not exactly kind
of cocktail dresses.

86
00:04:33,730 --> 00:04:38,030
So let me just click on this one and
find a visually similar cocktail dress.

87
00:04:39,160 --> 00:04:43,250
Now what you see is a bunch of
cocktail dresses that tend to have

88
00:04:43,250 --> 00:04:45,660
a bit more of interesting
patterns associated with them.

89
00:04:46,930 --> 00:04:52,850
So they're more interesting, more along
the lines of what she was describing.

90
00:04:52,850 --> 00:04:56,750
So if I scroll through this,
say okay, this look pretty cool.

91
00:04:56,750 --> 00:05:00,448
Maybe this one in the center, which is
blue, maybe it has some color to it,

92
00:05:00,448 --> 00:05:01,877
is what she's looking for.

93
00:05:01,877 --> 00:05:03,369
I click on that.

94
00:05:03,369 --> 00:05:05,020
I figure out, you know what?

95
00:05:06,350 --> 00:05:12,540
It's not formal enough for
a cocktail dress but if I scan down,

96
00:05:12,540 --> 00:05:17,830
you'll see that I can find a dress like
this one where it's a cocktail dress,

97
00:05:17,830 --> 00:05:23,410
it's formal enough it has a touch of
color to it and it's interesting.

98
00:05:23,410 --> 00:05:27,270
And it might be something that she'll
actually be interested in wearing.

99
00:05:28,478 --> 00:05:32,370
So here, we've see an intelligent
application that looks at text data,

100
00:05:32,370 --> 00:05:35,970
image data, uses deep learning, and does
some really cool and interesting things.

101
00:05:37,230 --> 00:05:41,740
So we saw this demo of deep learning for
visual product recommendation.

102
00:05:41,740 --> 00:05:42,570
But in your capstone,

103
00:05:42,570 --> 00:05:46,440
you do something really even cooler,
much more interesting than this.

104
00:05:46,440 --> 00:05:49,024
You're gonna take your
capstone project and

105
00:05:49,024 --> 00:05:51,679
you're going to combine
a variety of things.

106
00:05:51,679 --> 00:05:54,843
You combine recommenders
with a recommender a system.

107
00:05:54,843 --> 00:05:59,086
You're gonna do some text analysis,
sentiment analysis, what people are saying

108
00:05:59,086 --> 00:06:03,190
teviews about different products and
extract those reviews and analyze them.

109
00:06:03,190 --> 00:06:06,800
You're going to do some computer
vision to visually understand images.

110
00:06:06,800 --> 00:06:10,165
You're going to do some deep learning
to really take that computer vision

111
00:06:10,165 --> 00:06:12,040
techniques and
make them extremely accurate.

112
00:06:12,040 --> 00:06:13,490
And then you're gonna take it and

113
00:06:13,490 --> 00:06:16,790
deploy it like a web service like
the one I just showed you, that webpage.

114
00:06:16,790 --> 00:06:19,868
You're gonna have your very own
intelligent web service for

115
00:06:19,868 --> 00:06:23,789
products which you can interact with,
do some really interesting things and

116
00:06:23,789 --> 00:06:27,909
really show to a ton of people how you
created intelligence behind that service.

117
00:06:27,909 --> 00:06:31,739
[MUSIC]