1
00:00:00,330 --> 00:00:05,580
Okay now we've got some of the foundations of num pi down pat.

2
00:00:05,580 --> 00:00:06,340
Or maybe not.

3
00:00:06,340 --> 00:00:06,740
It's all right.

4
00:00:06,750 --> 00:00:07,980
You don't have to know anything up by heart.

5
00:00:07,980 --> 00:00:13,320
For now let's look at a little practical example to finish things off.

6
00:00:13,320 --> 00:00:18,530
Practical example now I'm pi in action.

7
00:00:18,750 --> 00:00:22,030
Play of the beautiful action music a beautiful wrong word.

8
00:00:22,050 --> 00:00:22,670
Let's go.

9
00:00:22,680 --> 00:00:23,880
Intense action music.

10
00:00:23,880 --> 00:00:25,470
Rock and roll.

11
00:00:25,500 --> 00:00:30,780
Now if we come back here to our Jupiter dashboard I've got a little folder here called images which

12
00:00:30,780 --> 00:00:32,060
I've just created.

13
00:00:32,160 --> 00:00:33,710
Now got some photos in here.

14
00:00:33,960 --> 00:00:34,880
So the whole premise.

15
00:00:34,890 --> 00:00:38,790
Remember we talked about it right at the start with none pi is in machine learning.

16
00:00:38,790 --> 00:00:40,950
You take whatever data you have.

17
00:00:41,010 --> 00:00:42,860
You turn it into numbers.

18
00:00:42,930 --> 00:00:46,780
You use a machine learning algorithm to find patterns in those numbers.

19
00:00:46,800 --> 00:00:52,710
So what if your data happened to be images and you wanted to find out information about those images.

20
00:00:52,720 --> 00:00:56,510
I'm going to create a little markdown cell here and import.

21
00:00:56,760 --> 00:01:00,120
I've got a panda image that we've used in the pandas section.

22
00:01:00,510 --> 00:01:05,030
So we're going to turn our wonderful panda we can see it there.

23
00:01:05,340 --> 00:01:06,050
Stoat.

24
00:01:06,090 --> 00:01:06,940
Give it a high five.

25
00:01:07,010 --> 00:01:17,200
I'm high five in the screen what we want to do is we want to turn an image into a non pi array because

26
00:01:17,320 --> 00:01:22,630
once we've turned this image into an umpire Ray then we can use the functions that we've learnt throughout

27
00:01:22,630 --> 00:01:26,380
this entire section to find patterns in that we can manipulate it a bit.

28
00:01:26,440 --> 00:01:28,800
We could pass it into a machine learning algorithm.

29
00:01:28,900 --> 00:01:31,150
The world is our oyster when our data is in numbers.

30
00:01:31,150 --> 00:01:37,330
So we have here mapped Gottlieb we're going to see map pop live in another section import in red.

31
00:01:37,330 --> 00:01:43,560
So this is just a little little package within map plot lib called M Reed which stands for image read.

32
00:01:43,660 --> 00:01:49,950
This is going to do is m read might use our shift tab to figure this out.

33
00:01:50,290 --> 00:01:52,780
Or is it going to be there.

34
00:01:52,780 --> 00:01:54,610
Maybe because we haven't imported it yet.

35
00:01:56,740 --> 00:01:58,550
We'll need something that's okay.

36
00:01:58,570 --> 00:02:01,090
We want to go shift tab.

37
00:02:01,090 --> 00:02:04,660
There we go read an image from a file into an array.

38
00:02:04,810 --> 00:02:09,890
That is exactly what we want today f name stands for file name I believe.

39
00:02:10,000 --> 00:02:12,670
String or file like the image file to read.

40
00:02:12,670 --> 00:02:14,100
Wonderful.

41
00:02:14,140 --> 00:02:23,510
Now Pan that image is in the file images under pan and up Pan G so we can pass it a string here.

42
00:02:23,820 --> 00:02:26,680
Panda dot P. J.

43
00:02:26,880 --> 00:02:34,580
What we might do is before we even look at it let's check what type it comes in panda we forgot the

44
00:02:34,580 --> 00:02:35,820
images folder.

45
00:02:35,900 --> 00:02:36,420
That's right.

46
00:02:37,860 --> 00:02:40,290
Class num pi and you're right.

47
00:02:40,800 --> 00:02:41,500
Yes.

48
00:02:41,500 --> 00:02:42,960
Now what does that mean.

49
00:02:42,960 --> 00:02:49,200
Well it means because it's an empty array format we can use all of the functionality that we've learned

50
00:02:49,200 --> 00:02:51,220
about num PI on this array.

51
00:02:51,240 --> 00:02:52,830
So let's have a look at it.

52
00:02:52,950 --> 00:02:57,510
Panda wow is a lot of numbers there.

53
00:02:58,350 --> 00:03:01,250
Let's check out some of the attributes of its size.

54
00:03:01,320 --> 00:03:05,730
Panda don't shape panda don't.

55
00:03:06,780 --> 00:03:10,680
And dim This is pretty big.

56
00:03:11,160 --> 00:03:12,150
What's that number.

57
00:03:12,150 --> 00:03:15,670
Twenty four million four hundred and sixty five thousand.

58
00:03:15,720 --> 00:03:18,050
It's got a lot of dimensions here.

59
00:03:18,210 --> 00:03:22,950
It's a three dimensional array because the three dimensions are color channels and what what this array

60
00:03:22,950 --> 00:03:28,060
actually is let's just view the first 100 to these numbers here.

61
00:03:28,170 --> 00:03:30,120
Maybe that's even too many.

62
00:03:30,300 --> 00:03:36,350
They want to view the first five There we go first five.

63
00:03:36,410 --> 00:03:43,130
What this is done is a little function im read I m Reid however you want to pronounce it has taken this

64
00:03:43,130 --> 00:03:49,580
picture and then found all of the little pixel value so if we zoomed right in here you might not even

65
00:03:49,580 --> 00:03:52,680
be to see a single pixel That's right.

66
00:03:52,700 --> 00:04:00,200
We zoomed right in here got the pixels those pixels have color values so red green and blue and that

67
00:04:00,260 --> 00:04:02,750
is what is stored in disarray.

68
00:04:02,750 --> 00:04:06,760
So if we did the same let's have a look maybe for another image.

69
00:04:06,860 --> 00:04:11,290
So we've got image source.

70
00:04:11,330 --> 00:04:17,390
We have another one images tab for auto complete car photo.

71
00:04:17,390 --> 00:04:18,190
This is what I have now.

72
00:04:18,200 --> 00:04:24,440
You can try this with your own image beautiful.

73
00:04:24,510 --> 00:04:34,320
So what we're gonna do is you're going to a car equals M read that same images car photo dot P G wonderful.

74
00:04:34,500 --> 00:04:39,120
We're going to check it again just for completeness type car.

75
00:04:40,340 --> 00:04:48,770
Now apply Ray beautiful so now we've turned this car photo into an endless array.

76
00:04:48,920 --> 00:04:50,410
Wonderful.

77
00:04:50,410 --> 00:04:51,250
Okay.

78
00:04:51,310 --> 00:04:58,400
I you just only see the first section of that let's have a look at one more photo to.

79
00:04:58,470 --> 00:05:04,590
It goes in read the images we've seen this dog photo before right at the start.

80
00:05:04,590 --> 00:05:06,150
My beautiful poppies.

81
00:05:06,360 --> 00:05:08,730
We're gonna turn them into numbers.

82
00:05:08,800 --> 00:05:10,540
Actually we want to view it first.

83
00:05:10,630 --> 00:05:18,970
So we'll go to source equals images dog photo if you have a pen.

84
00:05:18,970 --> 00:05:25,360
Image this raid function will work with with urine so you could try that to start using it.

85
00:05:25,360 --> 00:05:29,780
You just need to run this for Matt pop live in Port and Reed then you can use this.

86
00:05:29,860 --> 00:05:36,150
Remember to use shift tab if you want to figure out what it does turn this into marked down there's

87
00:05:36,160 --> 00:05:38,790
my beautiful poppies that's 7 that's Bella.

88
00:05:39,040 --> 00:05:47,160
So we're gonna turn the poppies into can you guess what data type num pi array.

89
00:05:47,240 --> 00:05:48,440
Wonderful.

90
00:05:48,440 --> 00:05:55,100
So I've got dog so now what we could do with this is we've done one of the biggest steps of all machine

91
00:05:55,100 --> 00:05:59,410
learning as we turned our data into numbers specifically a num pi array.

92
00:05:59,690 --> 00:06:04,080
So what a machine learning algorithm would do is because it can't really look at this.

93
00:06:04,280 --> 00:06:10,470
Ask people we can look at this and go that's a tree as a tree that the dog that's AIS that's a shelter.

94
00:06:11,240 --> 00:06:14,530
But what a machine learning algorithm needs is numbers.

95
00:06:14,690 --> 00:06:17,740
And that's what it's got here in the form of an umpire Ray.

96
00:06:17,780 --> 00:06:24,140
So what it would do is it could take say this tree is is this section of numbers here.

97
00:06:24,140 --> 00:06:30,260
Probably not but let's just imagine it is what it will do is then go to this image and then look at

98
00:06:30,590 --> 00:06:32,720
this section of numbers here and go Oh OK.

99
00:06:32,720 --> 00:06:34,430
Those numbers are pretty close to each other.

100
00:06:34,760 --> 00:06:41,830
So this item here in this image must be a tree and this must be some grass and then same thing in here.

101
00:06:41,840 --> 00:06:50,590
This must be some grass because the pixel values are very close together in both images so that's going

102
00:06:50,590 --> 00:06:52,780
to wrap up the NUM pi section.

103
00:06:52,780 --> 00:06:58,600
What the important takeaways from this is to remember that we have covered a lot and you don't need

104
00:06:58,600 --> 00:07:01,120
to know all of this off by heart to begin with.

105
00:07:01,120 --> 00:07:07,000
It's just I've gone through some of the major concepts and functionality of the name pi library.

106
00:07:07,000 --> 00:07:14,410
The biggest thing is that num pi handles data types like this arrays very very well.

107
00:07:14,410 --> 00:07:19,000
So if you're working with machine learning and you're working with doing mathematical operations you're

108
00:07:19,000 --> 00:07:23,540
trying to find patterns in data you're probably going to use num Pi.

109
00:07:24,100 --> 00:07:26,770
And if you don't understand everything right now that's okay.

110
00:07:26,770 --> 00:07:29,080
There's a lot to take him have a break.

111
00:07:29,080 --> 00:07:31,040
Play around with some of your own images.

112
00:07:31,060 --> 00:07:36,580
Try and turn them into num pi arrays go back through what we've covered and I'll see you in the next

113
00:07:36,580 --> 00:07:37,000
section.