1
00:00:00,450 --> 00:00:11,620
OK so last video we got in our labels dot CSB which is now in the data frame labels say it's vague and

2
00:00:11,620 --> 00:00:17,980
it contains image ideas which are file names and that their brave label and then we checked out the

3
00:00:17,980 --> 00:00:25,010
distribution of our breed labels and we figured out there's about 80 to also images per class.

4
00:00:25,210 --> 00:00:32,350
Now it'd be good if we could view an image inside the notebook because this is part of our exploratory

5
00:00:32,350 --> 00:00:35,110
data analysis is checking out our data.

6
00:00:35,140 --> 00:00:38,380
Let's do that let's you an image.

7
00:00:38,890 --> 00:00:42,550
So one way to do that is using Python.

8
00:00:42,550 --> 00:00:47,380
Remember even though we're working in Google lab this is still under the hood.

9
00:00:47,400 --> 00:00:50,980
You put a notebook so we can go from Python.

10
00:00:51,030 --> 00:00:56,080
Don't display import Image.

11
00:00:56,220 --> 00:01:06,160
Then we can write image and then we just pass it a file name and a little trick here in collab oh is

12
00:01:06,970 --> 00:01:11,950
before we had to press shift and TAB energy for notebook to find the doctoring in collab and me it's

13
00:01:11,950 --> 00:01:16,320
command space shift and then Siri comes up.

14
00:01:16,320 --> 00:01:19,590
So command shift space.

15
00:01:19,660 --> 00:01:23,910
There we go on Windows it might be control instead of command.

16
00:01:24,880 --> 00:01:25,420
So there we go.

17
00:01:25,420 --> 00:01:27,330
This is our doctoring here.

18
00:01:27,400 --> 00:01:32,560
We need to pass it some sort of data and if we want to see it we need to pass it a file name.

19
00:01:32,650 --> 00:01:36,730
Let's grab one of these image I.D. so I'll just copy that.

20
00:01:36,730 --> 00:01:37,630
It should be.

21
00:01:37,630 --> 00:01:43,880
Actually let's have a look at a dingo See I'm from Australia and we're pretty famous for a story called

22
00:01:43,880 --> 00:01:46,380
dingo took my baby.

23
00:01:47,360 --> 00:01:50,110
I won't go into the details of that but you can look it up.

24
00:01:50,390 --> 00:01:52,390
And then if we go here we go dry.

25
00:01:52,390 --> 00:01:53,790
We need a positive file name.

26
00:01:53,810 --> 00:01:57,190
My drive file path sorry.

27
00:01:57,230 --> 00:01:58,440
And then we go now.

28
00:01:58,790 --> 00:02:03,580
Dog vision folder dog vision slash.

29
00:02:04,080 --> 00:02:06,610
It's going to be in the training folder.

30
00:02:06,890 --> 00:02:14,430
This and we need to put on the end j peg because if we come up to Kaggle we can see that is actually

31
00:02:14,430 --> 00:02:16,040
what the phone name is.

32
00:02:16,050 --> 00:02:20,630
That's the I.D. but the full phone name has the Dodge IPG.

33
00:02:20,910 --> 00:02:28,500
Now let's close this bad boy and shift in into oh look at that cute little Dingo.

34
00:02:28,920 --> 00:02:31,470
But apparently they can be savages at times.

35
00:02:31,470 --> 00:02:33,710
That's one way to view images in our notebook.

36
00:02:33,900 --> 00:02:39,540
So if we needed to explore what image we're working with while we're turning it into a tensor we can

37
00:02:39,540 --> 00:02:46,560
use this little function here or I think there's another one with Matt Lib we can use in the show but

38
00:02:46,560 --> 00:02:47,700
we'll have a look at that one later.

39
00:02:48,660 --> 00:02:53,910
Now we can view an image and we've got the Firefly days and their braids.

40
00:02:53,910 --> 00:03:00,420
Let's get a list of all the file names so we don't have to type out this every single time because that

41
00:03:00,420 --> 00:03:01,730
would get annoying.

42
00:03:01,740 --> 00:03:05,280
So what we might do is write a little mock down cell here.

43
00:03:05,310 --> 00:03:15,930
So press command M.M. changed that into markdown and go here getting images and their labels.

44
00:03:16,260 --> 00:03:24,440
Let's get a list of all of our image file path names.

45
00:03:24,750 --> 00:03:25,910
Beautiful.

46
00:03:26,310 --> 00:03:27,660
And so how can we do this.

47
00:03:27,680 --> 00:03:38,310
So we might go here create path names from Image I.D. Now we've just done one here.

48
00:03:38,310 --> 00:03:39,150
So this is a path.

49
00:03:39,150 --> 00:03:46,380
Name of this particular image and we did it by stealing this value here from label so we might bring

50
00:03:46,380 --> 00:03:56,800
labels CSB down here just so we can have a gaze at it while we work with this label CSA dot head.

51
00:03:56,900 --> 00:03:57,500
There we go.

52
00:03:57,980 --> 00:04:05,600
So we might create a little variable to loop through this and create these long strings at the same

53
00:04:05,600 --> 00:04:06,440
time.

54
00:04:06,540 --> 00:04:12,410
So let's see if we can do that with a list comprehension so file names equals let's just go f name for

55
00:04:12,440 --> 00:04:14,420
F name to start with.

56
00:04:14,420 --> 00:04:21,600
We'll start small in labels CSA idea what does this return

57
00:04:24,380 --> 00:04:29,870
check the first 10 phone aims.

58
00:04:30,050 --> 00:04:31,310
Wonderful.

59
00:04:31,340 --> 00:04:33,910
So this is just a short little list comprehension.

60
00:04:33,920 --> 00:04:42,750
Basically it's saying it's create a list of f name for F name which is short file name in labels CSP

61
00:04:43,160 --> 00:04:44,150
idea column.

62
00:04:44,300 --> 00:04:50,900
So take this idea column go through each one of them and take the value and save it to the list.

63
00:04:50,900 --> 00:04:51,910
So that's what we've got there.

64
00:04:54,900 --> 00:05:01,600
Now we need to change it up to be a bit more like this because these at the moment are just ideas.

65
00:05:01,620 --> 00:05:14,970
So what we'll do is drive my drive dog vision trying as we can to drive slash my drive slash dog vision

66
00:05:15,840 --> 00:05:24,150
slash train and then it's gonna be plus f name on a and destroying plus f name.

67
00:05:24,150 --> 00:05:30,380
What does that return I said check the first 10 here but I've lied.

68
00:05:30,380 --> 00:05:32,060
I didn't put the little slice

69
00:05:36,780 --> 00:05:37,380
OK.

70
00:05:37,380 --> 00:05:44,170
And now we just need to add of course the file extension because otherwise it wouldn't work.

71
00:05:44,180 --> 00:05:48,540
These are GI pigs don't JP How does this look.

72
00:05:50,910 --> 00:05:52,210
Wonderful.

73
00:05:52,230 --> 00:05:52,930
OK.

74
00:05:53,250 --> 00:05:59,580
So now we've got a list of all the file names from the I.D. column of labels CSB what we should do is

75
00:05:59,580 --> 00:06:05,280
compare them to the number of files in our training data directory.

76
00:06:05,280 --> 00:06:06,450
Why would we do this.

77
00:06:06,450 --> 00:06:13,080
Well this is to make sure that we've got the same amount of file names as we do actual files in our

78
00:06:13,080 --> 00:06:14,220
training file.

79
00:06:14,250 --> 00:06:21,370
We downloaded this data from Kaggle we uploaded it to collaborate more specifically our dog vision folder

80
00:06:22,210 --> 00:06:23,390
and then we unzipped it.

81
00:06:23,390 --> 00:06:28,110
So we want to make sure that we're working with the same amount of data.

82
00:06:28,190 --> 00:06:28,940
So let's do that.

83
00:06:28,970 --> 00:06:41,150
So we go here check whether number of file names matches the number of actual image files because otherwise

84
00:06:41,300 --> 00:06:47,080
if we just kept going now and we've worked out later on that hey we've got some issue with our data

85
00:06:48,190 --> 00:06:51,220
then it's kind of going to undo a lot of the stuff that we've done.

86
00:06:51,400 --> 00:06:51,960
So that's it.

87
00:06:51,970 --> 00:06:59,620
That's a big point is a lot of machine learning is just basically massaging your data into a form so

88
00:06:59,620 --> 00:07:02,620
that it works with machine learning if that makes sense.

89
00:07:02,620 --> 00:07:07,720
So that's what we're doing now we're just verifying that the data we've got is the right format the

90
00:07:07,720 --> 00:07:15,460
right amount and then we know once we run our machine learning models they'll be all hunky dory or at

91
00:07:15,460 --> 00:07:17,520
least hunky dory as possible.

92
00:07:17,770 --> 00:07:25,280
So we're going to input ours and we can check the amount of files in a particular file by using list.

93
00:07:25,450 --> 00:07:27,920
So Len OS Lister.

94
00:07:28,480 --> 00:07:31,300
I'll just write the code and then we'll explain it

95
00:07:34,720 --> 00:07:50,280
train is that equal to land of file names and if it is print phone aims match actual amount of files

96
00:07:52,820 --> 00:07:54,120
proceed.

97
00:07:54,310 --> 00:08:05,090
It's like we're giving ourselves a little instruction in print and phone names do not match actual amount

98
00:08:05,090 --> 00:08:10,710
of files check the target directory.

99
00:08:10,750 --> 00:08:11,420
All right.

100
00:08:11,420 --> 00:08:15,540
So what this is going to do if you do our list list lets you see what this prints out

101
00:08:18,280 --> 00:08:19,410
code.

102
00:08:19,750 --> 00:08:25,900
Actually I might only do the first 10 again otherwise it'll be just spam name always is not defined.

103
00:08:25,900 --> 00:08:29,320
Oh that would make sense because of the import os there and not here.

104
00:08:30,340 --> 00:08:31,300
Let's do it up here.

105
00:08:34,680 --> 00:08:42,030
So what this is going to do is tell collab to hey go into dog vision train.

106
00:08:42,050 --> 00:08:44,390
Oh I've just called it data.

107
00:08:44,400 --> 00:08:45,860
There's no file there.

108
00:08:45,990 --> 00:08:48,680
Stop that.

109
00:08:48,890 --> 00:08:52,750
By the way if the executing code is not responding to interrupt.

110
00:08:52,760 --> 00:08:57,080
So if this comes up it's because this cell is unable to stop.

111
00:08:57,080 --> 00:08:59,700
So these are the things you run into.

112
00:08:59,910 --> 00:09:06,870
And the reason why this is going in sort of an infinite loop is because I actually do have a folder

113
00:09:06,870 --> 00:09:09,960
called data which is what this used to be.

114
00:09:09,960 --> 00:09:16,630
And within there there is also a try and follow so it's looping through there.

115
00:09:16,700 --> 00:09:17,530
What I'm gonna do.

116
00:09:17,630 --> 00:09:21,460
Can I can that then candidate.

117
00:09:21,460 --> 00:09:28,050
So let me just tell you what it does always Lister is just gonna go through a directory and then list

118
00:09:28,140 --> 00:09:29,580
all the files in that directory.

119
00:09:29,910 --> 00:09:36,180
So we're saying if the length of all the files in this drive my drive dark vision

120
00:09:39,780 --> 00:09:43,580
his our other warning that's popped up this is beautiful timing.

121
00:09:43,580 --> 00:09:46,460
So these are the actual things that you're going to run into.

122
00:09:46,520 --> 00:09:47,690
Get back to our keynote.

123
00:09:47,690 --> 00:09:48,510
Here we go.

124
00:09:49,480 --> 00:09:55,360
So this is a little area that you might experience if there's something going on with your google drive

125
00:09:55,360 --> 00:09:58,080
and the way that collab is interacting with it.

126
00:09:59,130 --> 00:10:02,310
So the good thing is it usually fixes itself.

127
00:10:02,310 --> 00:10:08,770
And if you want to put this more info tab you'll get this little paragraph of text.

128
00:10:08,770 --> 00:10:16,580
So basically it's saying if you have a lot of files in Google Drive in the top folder.

129
00:10:16,840 --> 00:10:24,800
So if we come back to dog vision so in this folder here if you have a lot of files there.

130
00:10:24,820 --> 00:10:30,610
Go drive can time out as well as if you have a lot of files within a particular folder.

131
00:10:30,780 --> 00:10:32,430
It can also time out.

132
00:10:32,550 --> 00:10:37,230
So that's very handy that we saw that in real time because I don't want you to come across these warnings

133
00:10:37,230 --> 00:10:38,520
and not be sure of what to do.

134
00:10:38,520 --> 00:10:42,860
But if in doubt remember you can always just search this asset question.

135
00:10:42,870 --> 00:10:44,540
I'm sure someone will be out to help you out.

136
00:10:44,550 --> 00:10:46,100
If not I'll help you out.

137
00:10:46,110 --> 00:10:47,730
Click the links.

138
00:10:47,730 --> 00:10:51,790
There's a bunch of things you can try out let's just run the code.

139
00:10:51,870 --> 00:10:53,060
We've spoken about it for too long.

140
00:10:54,560 --> 00:10:56,420
File names match actual amount of files.

141
00:10:56,430 --> 00:10:57,390
Proceed.

142
00:10:57,390 --> 00:11:01,590
Now I will let you test this one out yourself.

143
00:11:01,590 --> 00:11:05,330
We're going to follow the orders of this code here and proceed.

144
00:11:05,390 --> 00:11:08,520
OK so everything's working.

145
00:11:08,520 --> 00:11:16,160
We have all the file names in a list and we have the same amount of file names as we have actual images.

146
00:11:16,170 --> 00:11:21,390
Let's do one more check to make sure that we're visualizing directly from a fall apart so we can go

147
00:11:21,390 --> 00:11:27,070
here one more check.

148
00:11:27,100 --> 00:11:30,370
Remember we want to make sure our data is in the right format.

149
00:11:30,820 --> 00:11:33,220
Let's go the nine thousand.

150
00:11:33,220 --> 00:11:39,690
Image nine thousand index.

151
00:11:39,860 --> 00:11:42,740
Good Lord.

152
00:11:43,190 --> 00:11:45,480
I wonder what type of dog that is.

153
00:11:45,500 --> 00:11:47,360
What an absolute base we can find out.

154
00:11:47,360 --> 00:11:47,780
Really.

155
00:11:48,230 --> 00:11:49,550
Let's go here.

156
00:11:50,120 --> 00:11:56,000
Labels CSB I'd say nine thousand

157
00:11:59,130 --> 00:12:00,710
oh no we won't breed

158
00:12:04,510 --> 00:12:05,740
that thing looks Savage.

159
00:12:05,740 --> 00:12:09,030
Look at the size of that chain Tibetan mastiff.

160
00:12:09,250 --> 00:12:14,080
That even sounds like a tough dog I want one of them.

161
00:12:14,090 --> 00:12:16,370
That's a base that's like a dog Hercules would have.

162
00:12:16,550 --> 00:12:17,790
OK.

163
00:12:18,010 --> 00:12:20,220
Now we've got all of our phone names that are correct.

164
00:12:20,230 --> 00:12:22,710
What we're going to have to do so we've got the file names.

165
00:12:22,810 --> 00:12:25,170
Now we need to work with our labels.

166
00:12:25,180 --> 00:12:29,040
We need to get our labels in a format that we can use.

167
00:12:29,050 --> 00:12:31,020
So let's do that in the next video.