1
00:00:04,960 --> 00:00:08,670
In a previous lecture we haven't really created a brand new.

2
00:00:08,800 --> 00:00:15,910
I sure couldn't give surveys that is going to allow us to provide some computer vision analysis into

3
00:00:15,910 --> 00:00:18,090
our applications.

4
00:00:18,190 --> 00:00:26,260
We're going to start in these lecture by analyzing images that will be able to find in the Internet.

5
00:00:26,260 --> 00:00:33,160
So we're going to get yourself a certain image and we're going to send it to the custom Bishan API for

6
00:00:33,220 --> 00:00:42,710
its analysis of course first things first I'm here on my Asherah notebooks and in my LPA dash washerwomen

7
00:00:42,770 --> 00:00:48,230
library what I am going to do is create a brand new notebook.

8
00:00:48,510 --> 00:00:54,630
We are going to be getting back into the azure notebooks in the previous sections we have been working

9
00:00:54,630 --> 00:01:03,000
with local Jupiter because we required local functionalities such as access to the microphone in this

10
00:01:03,000 --> 00:01:05,750
section we're not going to be using the microphone anymore.

11
00:01:05,910 --> 00:01:09,970
So that is why I can be back here in your notebooks.

12
00:01:10,920 --> 00:01:15,520
Now my new notebook is simply going to be called Computer vision.

13
00:01:16,110 --> 00:01:21,430
And of course that type of fire is going to be applied in 3.6 notebook.

14
00:01:21,780 --> 00:01:30,060
So I'm going to hit on new and I'm going to be opening this notebook and at it now because in this notebook

15
00:01:30,060 --> 00:01:36,810
I'm going to be using that computer vision service that I have just created what I will do is get back

16
00:01:36,900 --> 00:01:40,700
to my actual portal where I have these computer vision.

17
00:01:40,750 --> 00:01:48,450
I sure service I'm going to get on one of the keys I'm going to create in the first cell of my new notebook

18
00:01:48,990 --> 00:01:56,280
a subscription key variable that I will simply use whenever I have to access the service.

19
00:01:56,360 --> 00:02:02,510
So have something like subscription King here which is going to be equal to that key that I just got

20
00:02:02,510 --> 00:02:05,030
it and that will just wrong that cell.

21
00:02:05,070 --> 00:02:08,500
So these variable is available now.

22
00:02:08,520 --> 00:02:16,530
The first thing that I'm going to be defining here to be able to connect to that service is pretty much

23
00:02:16,590 --> 00:02:25,050
every other time that we have worked the headers and the others are going to require only one value

24
00:02:25,530 --> 00:02:33,350
the same value that we have been using which is OCB that's A.P.M. that's prescription Dashti optionally

25
00:02:33,360 --> 00:02:39,300
of course we can always say the Content-Type and the accepte Heathers but I believe that by default

26
00:02:39,660 --> 00:02:42,350
both are set to Jasen which is what we want.

27
00:02:42,420 --> 00:02:48,900
If you ever see that that is not the case in any of these services you can simply sign it from the headers

28
00:02:49,230 --> 00:02:50,400
to application form.

29
00:02:50,400 --> 00:02:57,400
Plus Jason in this case I'm only going to be setting the OCP a subscription fee.

30
00:02:57,930 --> 00:03:02,380
And of course its value is going to be the subscription King.

31
00:03:02,820 --> 00:03:11,640
Next let me define the body and the body is going to require one single thing which is going to be in

32
00:03:11,640 --> 00:03:16,280
the case of what we are going to be trying here.

33
00:03:16,950 --> 00:03:20,640
So the body can be three different things.

34
00:03:20,700 --> 00:03:27,900
It can be adjacent which is going to be the case in this example or we can be an OC to the stream or

35
00:03:27,950 --> 00:03:34,170
our form data a multipart form data similar to what we used in the previous section when sending an

36
00:03:34,200 --> 00:03:35,850
audio file.

37
00:03:35,850 --> 00:03:43,640
Now the reason for these two other scenarios is because we can analyze in the mansion a local image

38
00:03:44,070 --> 00:03:51,230
instead of analyzing remote image and image that we are going to be asked to think through your arrival.

39
00:03:51,840 --> 00:03:54,570
So the creation of the body is going to be important.

40
00:03:54,690 --> 00:03:59,830
It is going to define what exactly you are going to be using in here.

41
00:04:00,150 --> 00:04:05,370
Either yourself for an image that you have already uploaded or for on the image that you found on the

42
00:04:05,370 --> 00:04:13,560
Internet or multipart or not to the stream with the information about the image that you have locally

43
00:04:14,700 --> 00:04:20,940
for this particular example then we're going to be working with our remote image using Jason in here

44
00:04:21,320 --> 00:04:24,360
sitting simply the value of the rail.

45
00:04:24,390 --> 00:04:31,080
For example in this case I said to the Eurail to be a certain image that I found on the web you can

46
00:04:31,080 --> 00:04:36,340
do a quick Google search or just found an image that you want to analyze.

47
00:04:36,510 --> 00:04:38,720
In this case I can give you.

48
00:04:38,790 --> 00:04:45,420
And this is going to be and you can actually guess what this is about because this has a name pretty

49
00:04:45,420 --> 00:04:48,130
giving name that says something about Paris.

50
00:04:48,270 --> 00:04:50,100
So this is a photo of Paris.

51
00:04:50,250 --> 00:04:56,430
Just make sure that when you said that you are rail and these ends on the extension of the Mishler wise

52
00:04:56,430 --> 00:04:59,460
it is not really an image you are real.

53
00:04:59,700 --> 00:05:08,760
Also make sure that your ear is ears and big ears BMG ears a gift or ease a BNP file.

54
00:05:08,850 --> 00:05:17,320
The image cannot be more than four megabytes and it has to be bigger than 50 pixels by 50 pixels.

55
00:05:17,640 --> 00:05:22,650
So use mixture of those two things and you will have your body ready.

56
00:05:22,650 --> 00:05:27,900
In addition to the ball you of course we're going to meet the parameters and the parameters are going

57
00:05:27,900 --> 00:05:31,510
to have three optional values.

58
00:05:31,650 --> 00:05:32,780
So limit the fine here.

59
00:05:32,790 --> 00:05:36,800
The parameters now to be able to define the Brehm's.

60
00:05:36,840 --> 00:05:42,390
I am going to need to import a couple of things that let me just import the usual thing here.

61
00:05:42,630 --> 00:05:46,540
I will start with your L-EB that Parr's.

62
00:05:46,750 --> 00:05:48,690
We will also need HTP.

63
00:05:48,710 --> 00:05:55,920
The client will need Jason and with this I shall be able to get the programs by calling from you r l

64
00:05:57,470 --> 00:06:06,390
lived Parr's they your L and Cosmas and these methods ees going to receive the three optional values

65
00:06:06,390 --> 00:06:09,950
that can be received by the service.

66
00:06:10,440 --> 00:06:11,840
So these are optional bulled.

67
00:06:11,880 --> 00:06:16,160
I'm going to mention them for you to keep in mind what you can setting here.

68
00:06:16,830 --> 00:06:24,240
So the first one is going to be the visual features and the visual features are variable or volume and

69
00:06:24,240 --> 00:06:29,810
here will be able to define what is it that you want the service to return to you.

70
00:06:30,360 --> 00:06:33,360
And here you can set many different values.

71
00:06:33,360 --> 00:06:35,960
This is going to be a string value.

72
00:06:36,090 --> 00:06:41,830
You can set if you want to return a check or is if you want the service to get some tags for the image.

73
00:06:42,000 --> 00:06:49,650
If you want a description if you want to identify faces which can be pretty useful in case you want

74
00:06:49,920 --> 00:06:59,280
to determine gender and age and even where the face is so it's quickly and easy to maybe draw as well

75
00:06:59,330 --> 00:07:02,080
around these days.

76
00:07:02,190 --> 00:07:11,120
You want an image type color and whether or not DC tagged Suomi mesh that should not be seen by kids.

77
00:07:11,500 --> 00:07:16,620
So we're going to do in here eat first and at least for this particular image.

78
00:07:16,620 --> 00:07:18,140
That is of parries.

79
00:07:18,390 --> 00:07:26,210
I am going to be asking for a description so I am going to say we show features to be description.

80
00:07:26,220 --> 00:07:35,820
Additionally we could get the details and the details is just another string of journalists always that

81
00:07:36,060 --> 00:07:43,680
is going to help us identify two things either celebrities or landmarks and it's simply going to be

82
00:07:43,800 --> 00:07:44,470
returning US.

83
00:07:44,490 --> 00:07:51,890
Whether or not it identifies any landmark or any celebrity in the picture for these particular one of

84
00:07:51,950 --> 00:07:56,340
parries we're going to try landmarks.

85
00:07:56,970 --> 00:08:03,810
And finally you could be saying the language the language in which you want the service to return to

86
00:08:03,810 --> 00:08:09,900
information in by default is English at the moment of recording this video.

87
00:08:09,930 --> 00:08:17,240
The only support languages are English DePrince is Portuguese and simplified Chinese.

88
00:08:17,250 --> 00:08:19,290
I am going to be leaving this in English.

89
00:08:19,290 --> 00:08:20,240
This is the default.

90
00:08:20,310 --> 00:08:26,460
So I wouldn't have to establish it but I'm just going to be sending it in you so you know that this

91
00:08:26,460 --> 00:08:33,350
is something that you can do and do everything that we need whether it's body parameters in the body

92
00:08:33,350 --> 00:08:34,220
is important.

93
00:08:34,230 --> 00:08:41,240
Think that you are a rail or if we have a local image the local image in the form of a multi-port were

94
00:08:41,240 --> 00:08:43,330
enough to stream.

95
00:08:43,410 --> 00:08:45,410
So now let's make the request.

96
00:08:45,780 --> 00:08:51,630
I am going to be doing it in a try catch block as pretty much every single time and the first thing

97
00:08:51,630 --> 00:08:59,880
is to create the connection which is going to be equal to each GTP the client got a GDP connection so

98
00:08:59,910 --> 00:09:09,230
HTP disconnection and the connection is going to be established to every aimpoint or at least that Eurail

99
00:09:09,250 --> 00:09:17,640
the weekend finding herefore Asher account more specifically inside of the overview for our computer

100
00:09:17,640 --> 00:09:25,850
vision service each year as we can see it is deal Daut API that cognitive Dolemite quoted com.

101
00:09:26,040 --> 00:09:32,560
But the important thing to notice in here ease the location of our service.

102
00:09:32,660 --> 00:09:35,240
In my case south central US.

103
00:09:35,330 --> 00:09:38,210
In your case and these may vary.

104
00:09:38,210 --> 00:09:46,400
So it is very important that you do said this first time of the Eurail to the location where you created

105
00:09:46,400 --> 00:09:47,870
the service.

106
00:09:48,230 --> 00:09:54,750
So that is what I am going to be passing to the HTP s connection constructor and with the connection

107
00:09:54,750 --> 00:09:57,470
I established I can make the request.

108
00:09:57,950 --> 00:10:01,570
The request is going to be a post request.

109
00:10:01,820 --> 00:10:05,800
The point I havent created in the year board.

110
00:10:05,810 --> 00:10:13,580
If we never get back to usher you can see how by the way this is deal 1.0 as we noticed in the previews

111
00:10:13,580 --> 00:10:14,330
lecture.

112
00:10:14,570 --> 00:10:20,290
But it's definitely still going to be for less Rhaetian for flash and the version.

113
00:10:20,450 --> 00:10:24,280
In our case we're already going to be setting that to 2.0.

114
00:10:24,510 --> 00:10:30,460
So that is going to be for Flash vision for Flash VI 2.0.

115
00:10:30,470 --> 00:10:37,940
In our case and to get this particular point we're going to be adding a Forth less and analyze.

116
00:10:38,090 --> 00:10:44,660
So as I mentioned in a previous lecture the computer Bishan API is going to have many different endpoints

117
00:10:45,050 --> 00:10:46,120
who will be able.

118
00:10:46,120 --> 00:10:56,090
Among other things to make some analysis of the images to do some OCR to identify handwritten text etc..

119
00:10:56,570 --> 00:11:02,650
For these particular servers that we're using right now the End Point is analyze.

120
00:11:03,050 --> 00:11:09,140
And if we're going to be analyzing we're going to be requiring a couple of Purab since you're the ones

121
00:11:09,140 --> 00:11:11,830
I already have in the parabens value.

122
00:11:12,110 --> 00:11:18,640
So I am going to be passing the parameters so this gets added to the end of the endpoint.

123
00:11:18,650 --> 00:11:22,230
Next I am going to be sending the body as a string.

124
00:11:22,910 --> 00:11:24,690
This is going to be Jason.

125
00:11:25,220 --> 00:11:28,650
And finally I am going to be sending the headers and that's it.

126
00:11:28,640 --> 00:11:34,020
I shall be able to get the response by calling that good response.

127
00:11:34,110 --> 00:11:38,100
Get some Jason data by calling from the responses.

128
00:11:38,300 --> 00:11:39,500
The wreathe method.

129
00:11:39,920 --> 00:11:43,840
And finally get the data itself by calling from Jason.

130
00:11:44,030 --> 00:11:53,220
The load method and passing that Jason data has always do NOT forget to close the connection and finally

131
00:11:53,280 --> 00:11:56,790
I shall catch any exception that may happen.

132
00:11:57,600 --> 00:11:59,530
So there is everything that we need.

133
00:11:59,910 --> 00:12:05,330
Let's test this out remember that the men in these games is a photo of Paris.

134
00:12:05,460 --> 00:12:10,430
In fact before doing anything let me show you what this image is about.

135
00:12:10,540 --> 00:12:13,890
I am going to here in that particular you are l.

136
00:12:14,060 --> 00:12:16,020
You can see this is a simple photo.

137
00:12:16,020 --> 00:12:21,630
There is a Boer's right here but you can see the Eiffel Tower so it's pretty easy to understand that

138
00:12:21,630 --> 00:12:23,550
this is Parrys.

139
00:12:23,550 --> 00:12:30,610
So now that you have seen that image let's run this el hand of course I forgot something very important

140
00:12:31,000 --> 00:12:33,640
which is to print the data.

141
00:12:33,790 --> 00:12:41,780
Let me try one more time and now we see these days in these digital string is going to contain some

142
00:12:41,780 --> 00:12:44,180
information about the Amish.

143
00:12:44,480 --> 00:12:51,380
Now taking a look at these two string you can see how it is identifying the boss as the main part of

144
00:12:51,380 --> 00:12:52,540
the photo.

145
00:12:52,820 --> 00:12:55,570
At least in some of the places inside here.

146
00:12:55,580 --> 00:12:57,270
For example forget Greece.

147
00:12:57,560 --> 00:13:02,120
We do see that one of the categories is a building.

148
00:13:02,380 --> 00:13:09,610
And one of the categories ease our door to these two categories have been identified.

149
00:13:09,680 --> 00:13:15,170
There is also a description which is going to have a first some tags it's correctly taking these hest

150
00:13:15,170 --> 00:13:19,220
outdoors road OK Hassam callers.

151
00:13:19,400 --> 00:13:26,870
It identifies a boss instead of the description it's also defining a caption which is going to be assigning

152
00:13:26,870 --> 00:13:33,800
it to exit that as you can see is entirely focused on the box it says this a double decker boss parked

153
00:13:33,800 --> 00:13:34,730
on the side of a road.

154
00:13:34,730 --> 00:13:40,200
So it's not really identifying anything about parries or the Eiffel Tower.

155
00:13:40,480 --> 00:13:46,200
It easier to find the height and the width and the formant correctly which is pretty easy.

156
00:13:46,390 --> 00:13:51,750
Well you can see how we can make these requests to the service.

157
00:13:51,860 --> 00:13:58,810
Now let me change this image to something now that contains people and what I'm going to do now is in

158
00:13:58,880 --> 00:14:06,260
the visual features instead of description I'm going to be asking for places so we can take a look at

159
00:14:06,260 --> 00:14:09,330
what this service is going to be returning.

160
00:14:09,850 --> 00:14:11,570
I am going to be changing them.

161
00:14:11,570 --> 00:14:13,170
There you are Ralph for the body.

162
00:14:13,200 --> 00:14:18,800
I have the image right here making sure again that this is a jaybirds or at least has the extension

163
00:14:19,250 --> 00:14:25,850
of any machine and changing the visual features two phrases normally show you the image and this is

164
00:14:25,880 --> 00:14:32,360
probably going to be old to have on the service because it contains a lot of face to see here.

165
00:14:32,570 --> 00:14:35,270
And they aren't precisely behind out of the picture.

166
00:14:35,270 --> 00:14:37,020
They are they were far away.

167
00:14:37,370 --> 00:14:40,420
Let's take a look at how the service behaves.

168
00:14:40,590 --> 00:14:42,870
I am going to run this one more time.

169
00:14:44,360 --> 00:14:50,620
And now in the Jason we see that their ears and faces are re being returned.

170
00:14:51,030 --> 00:14:57,880
Unfortunately for these image it was a bit hard for the service to identify them to returns and M-theory

171
00:14:59,010 --> 00:15:04,940
but you can definitely see that dignitaries have identified a crowd of people.

172
00:15:04,980 --> 00:15:07,980
We see a people group as well.

173
00:15:08,080 --> 00:15:11,590
You can see now the difference between passing faces.

174
00:15:11,590 --> 00:15:17,130
Ampacity other scription we see definitely way less information special Hindis is in there.

175
00:15:17,130 --> 00:15:20,120
We're phrases we're not identified.

176
00:15:20,230 --> 00:15:25,870
Let me change these Eurail to an image dad has way less faces.

177
00:15:25,920 --> 00:15:28,810
These may be easier for the service to work with.

178
00:15:29,700 --> 00:15:36,840
So I am going to be sitting there Eurail to a brand new Umich again making sure that these has that

179
00:15:36,880 --> 00:15:38,020
image extension.

180
00:15:38,100 --> 00:15:39,390
Let me show you the image right here.

181
00:15:39,390 --> 00:15:44,010
This is Mr. President Trump and the Queen of England.

182
00:15:44,040 --> 00:15:45,670
So this is way easier.

183
00:15:45,690 --> 00:15:48,930
Only three faces and pretty big inside of the image.

184
00:15:48,930 --> 00:15:52,920
So let's take a look how the service behaves in this scenario.

185
00:15:53,160 --> 00:15:55,470
I haven't changed any of the parameters.

186
00:15:55,570 --> 00:16:01,890
These business deals setting their faces as we show features and we now see how the interfaces we have

187
00:16:01,890 --> 00:16:03,700
in the rate of elements.

188
00:16:03,870 --> 00:16:12,450
The first one is going to be defining gender its identifying the Queen as very young so any one pretty

189
00:16:12,450 --> 00:16:12,810
close.

190
00:16:12,810 --> 00:16:19,200
I think shes about 20 years older and we see the phase recked and will which is not only setting all

191
00:16:19,650 --> 00:16:21,560
the coordinates.

192
00:16:21,630 --> 00:16:29,790
A one eighty one and two hundred and five but its also turning all the weed and the height which would

193
00:16:29,790 --> 00:16:36,090
be pretty easy for all to draw these squares out of image.

194
00:16:36,360 --> 00:16:42,990
Then it is identifying another E-Machine here which is the second them final phase that its detecting

195
00:16:42,990 --> 00:16:45,210
zero its not detecting the guard's face.

196
00:16:45,210 --> 00:16:53,440
Probably because it is half covered but its correctly identifying the gender and its returning are phased

197
00:16:53,440 --> 00:17:01,260
Rixton will very easily so you can see how in something easier for that or is it used correctly returning

198
00:17:01,530 --> 00:17:03,220
an array of phrases.

199
00:17:03,660 --> 00:17:07,400
So this is how you could be identifying this now in detail.

200
00:17:07,410 --> 00:17:14,080
We are currently doing these two landmarks we could actually change this to be celebrities.

201
00:17:14,790 --> 00:17:17,190
There are a couple of celebrities in this picture.

202
00:17:17,190 --> 00:17:19,850
Maybe this is going to identify them.

203
00:17:20,060 --> 00:17:21,410
Border want to know what is.

204
00:17:21,430 --> 00:17:28,210
Here is how changing the visual features he's going to return you the information that you want.

205
00:17:28,320 --> 00:17:34,190
You can see this two tags two categories two description one that I think is quite important is getting

206
00:17:34,200 --> 00:17:35,630
these two adult.

207
00:17:35,870 --> 00:17:43,260
If you said this to a told you will be able to identify a certain image has pornography in it and that

208
00:17:43,260 --> 00:17:46,280
information may be banished from your side.

209
00:17:46,530 --> 00:17:50,420
Maybe prevent this from being displayed to young kids.

210
00:17:50,550 --> 00:17:58,090
So you can start to add certain functionality to your application so it is more suitable for children.

211
00:17:58,650 --> 00:18:05,750
Let's for now does this with the same image but changing the details from Larrimer to celebrities.

212
00:18:06,060 --> 00:18:07,670
Let me run this.

213
00:18:07,670 --> 00:18:14,060
And now you can see that in the details that we have right here we're going to be having celebrities.

214
00:18:14,250 --> 00:18:17,010
First there is going to be ashtrays rectangle.

215
00:18:17,010 --> 00:18:18,810
It's still going to be the most.

216
00:18:18,810 --> 00:18:21,270
Where is this person.

217
00:18:21,280 --> 00:18:27,270
And you can see that now we see the name which is Elizabeth the Second correctly identified I'm pretty

218
00:18:27,270 --> 00:18:31,740
confident at ninety nine point ninety eight percent.

219
00:18:31,830 --> 00:18:33,910
Then there is another face rectangle.

220
00:18:33,910 --> 00:18:41,910
It also tells us where the location of this face is the width and height and the name Donald Trump.

221
00:18:41,920 --> 00:18:50,670
Again very confident at ninety nine point sixty two percent and eventually psaltery turning that faces

222
00:18:50,780 --> 00:18:52,100
the same Gordon.

223
00:18:52,140 --> 00:18:57,920
But remember that the faces a difference because it also tells us something about the age and gender

224
00:18:58,520 --> 00:19:02,830
in this case not very accurately identifying the age.

225
00:19:02,840 --> 00:19:07,250
Maybe they do look Giang but accurately identifying the gender.

226
00:19:07,460 --> 00:19:15,290
And of course where the faces are with this information in the same image you could start to maybe draw

227
00:19:15,290 --> 00:19:21,740
a rectangle around the faces maybe even grating the name of that people identify them.

228
00:19:21,770 --> 00:19:28,980
Eventually you could also use the age and the gender to tell some information about the people there.

229
00:19:29,030 --> 00:19:34,940
Now the purpose of these particular letters is not to draw all the squares in this picture we're going

230
00:19:34,940 --> 00:19:39,060
to be doing that in the next section when we work with the Faces AAPI.

231
00:19:39,320 --> 00:19:43,460
But it is something that you could do now before finishing this lecture.

232
00:19:43,460 --> 00:19:49,580
I do want to mention that these days is it's little more difficult to understand that the ones we have

233
00:19:49,580 --> 00:19:53,990
worked with before this has more information it is more cluttered.

234
00:19:54,230 --> 00:19:59,800
And in fact we see elements being repeated here we see of the second.

235
00:19:59,790 --> 00:20:02,590
And Donald Trump twice this year.

236
00:20:02,660 --> 00:20:03,790
Why is this.

237
00:20:03,980 --> 00:20:10,780
Well to be able to analyze it Jayson better instead of printing the data directly would get our form

238
00:20:10,780 --> 00:20:15,710
of the data by calling from Jason Dunham's function.

239
00:20:15,710 --> 00:20:21,800
This function is of course going to be requiring that Jason not it's going to format but it is also

240
00:20:21,800 --> 00:20:29,840
going to allow us to do a couple of things one that is not quite important is sorting the keys so we

241
00:20:29,840 --> 00:20:33,880
can set short keys to truth.

242
00:20:34,400 --> 00:20:44,030
This is simply going to be sorting the name of the elements object keys alphabetically for example detail

243
00:20:44,060 --> 00:20:49,310
is going to appear before name and name is going to appear before us Corps.

244
00:20:49,460 --> 00:20:52,790
It's not that important but it's something that you can do.

245
00:20:52,940 --> 00:20:59,180
The one that is going to be important is the indent value which is going to be sitting to do.

246
00:20:59,660 --> 00:21:06,680
Well this is going to do is had some indentation so it's easier for us to understand what good is part

247
00:21:06,680 --> 00:21:08,110
of what other key.

248
00:21:08,360 --> 00:21:14,490
So let's run this as you can see now the agent is much much easier to understand.

249
00:21:14,900 --> 00:21:19,850
We can see how the main object is going to have a Teguh risky.

250
00:21:19,850 --> 00:21:27,080
It's also going to have our faces key and made data and request ITC's.

251
00:21:27,430 --> 00:21:33,640
Now there are two important ones are the categories and the faces we can see how terrorist is an array

252
00:21:33,910 --> 00:21:37,580
that is going to have a couple of objects in here.

253
00:21:38,100 --> 00:21:39,710
One two.

254
00:21:39,970 --> 00:21:45,250
And it's also going to contained in the underscore of the category which is the important thing for

255
00:21:45,250 --> 00:21:46,640
that category.

256
00:21:46,870 --> 00:21:51,520
And we see how one is called people whip and the other one is people.

257
00:21:51,610 --> 00:21:54,440
And there is some score for each of them.

258
00:21:54,850 --> 00:22:02,020
But notice that it is instead of each category where the details are being added and because there are

259
00:22:02,020 --> 00:22:09,930
parameters we see the details to have celebrities we see celebrities and this is why we see it twice

260
00:22:09,940 --> 00:22:13,570
because each category is going to contain those details.

261
00:22:14,440 --> 00:22:16,000
As you can see here.

262
00:22:16,510 --> 00:22:23,560
So that is why we see a little of the second and Donald Trump twice because they are and these details

263
00:22:23,590 --> 00:22:25,790
are inside of each category.

264
00:22:26,110 --> 00:22:31,840
And of course instead of celebrities there is that array with it two people being identified in there

265
00:22:32,500 --> 00:22:37,010
the confidence the name and even the phrase race and all that we saw.

266
00:22:38,000 --> 00:22:44,750
And then in the faces key instead of the main object we're also going to have on the rate in this case

267
00:22:44,750 --> 00:22:51,980
with two elements these do not have information about who is in there but it does I did defy age and

268
00:22:51,980 --> 00:22:59,400
gender at this early and finally you know the myth that it requires that I mention.

269
00:22:59,620 --> 00:23:03,570
So this is how you can make the request of the computer vision API.

270
00:23:04,060 --> 00:23:10,960
In this particular scenario to make some analysis in the next lecture we were of course going to continue

271
00:23:10,960 --> 00:23:16,720
to work with the computer vision API but we're going to be working with another endpoint that is going

272
00:23:16,720 --> 00:23:18,930
to allow us to get a description.

273
00:23:19,270 --> 00:23:26,740
So it's easier for us to describe with our application what is happening is how the Amish many applications

274
00:23:26,780 --> 00:23:33,280
since you're the one that I have seen being apply the most used to describe an image to a blind person

275
00:23:33,700 --> 00:23:34,480
through audio.