1
00:00:00,050 --> 00:00:07,010
This platform permits us to curate better data more easily by helping us visualize, as you could see

2
00:00:07,010 --> 00:00:11,360
here, our data in a very easy and intuitive manner.

3
00:00:11,660 --> 00:00:17,600
Um, also helping us, um, as we said, already curate this, this high quality data sets, you have,

4
00:00:17,600 --> 00:00:24,110
um, public data sets or common data sets like the Coco, Open Images and Active Net, which are already

5
00:00:24,110 --> 00:00:24,980
available.

6
00:00:24,980 --> 00:00:34,010
Thanks to this 51 tool, we could easily find mistakes and this would help us, uh, train better models,

7
00:00:34,010 --> 00:00:38,300
as data quality is a key limiting factor on model performance.

8
00:00:38,300 --> 00:00:43,730
So, uh, the 51 tool permits you to easily identify these mistakes.

9
00:00:44,120 --> 00:00:49,790
We also have this, um, 51 brain, which helps you identify edge cases.

10
00:00:49,790 --> 00:00:54,590
You could check out this figure to the left where we have this visualization of embeddings.

11
00:00:54,590 --> 00:01:01,640
And as you may know, with embeddings or by visualizing embeddings in this way, we are able to get

12
00:01:01,640 --> 00:01:04,190
much more information about our data set.

13
00:01:04,190 --> 00:01:08,360
And this really plays a great role when curating data.

14
00:01:08,360 --> 00:01:16,430
Then finally, because we're able to curate and evaluate our data much faster, this leads to us or

15
00:01:16,430 --> 00:01:19,850
different teams getting to production even faster.

16
00:01:19,850 --> 00:01:30,530
So with this, we will dive into how we could make use of the 51 app to visualize, uh, segmentation

17
00:01:30,530 --> 00:01:32,570
or semantic segmentation data.

18
00:01:32,570 --> 00:01:38,240
Diving back into the code, we'll start by installing 51 and then importing it.

19
00:01:38,240 --> 00:01:42,890
So import 51 as F0.

20
00:01:42,920 --> 00:01:43,580
That's it.

21
00:01:43,580 --> 00:01:48,290
So uh, while this is installing we'll get into the user guide.

22
00:01:48,290 --> 00:01:53,360
So we have this documentations right here which you could feel free to check out.

23
00:01:53,360 --> 00:02:00,170
It's quite elaborate and contains um, a lot of stuff you would want to learn about the 51 app.

24
00:02:00,170 --> 00:02:07,880
So that said we could dive into first of all, you could check out the basics and then, um, dive into

25
00:02:07,880 --> 00:02:08,750
loading data sets.

26
00:02:08,780 --> 00:02:10,070
Let's get back to basics.

27
00:02:10,070 --> 00:02:10,250
Here.

28
00:02:10,250 --> 00:02:12,230
We have the basics here.

29
00:02:12,230 --> 00:02:15,950
You see how we have the key terms explained like data sets.

30
00:02:15,950 --> 00:02:17,630
We have samples.

31
00:02:17,630 --> 00:02:26,060
We have fields media types, tags, metadata labels, data set views and aggregations.

32
00:02:26,510 --> 00:02:29,570
Um nonetheless we're going to dive straight into loading the data sets.

33
00:02:29,570 --> 00:02:32,660
So we have loading data sets loading data sets from disk.

34
00:02:32,660 --> 00:02:39,770
Because we have our data set in our, uh, drive or in our colab notebook, which we've downloaded.

35
00:02:39,770 --> 00:02:42,770
So we could simply search for segmentation.

36
00:02:42,770 --> 00:02:43,970
Segmentation.

37
00:02:43,970 --> 00:02:47,600
You could see already that we have different types of data sets which you could load.

38
00:02:47,600 --> 00:02:50,570
We have object detection, we have image classification.

39
00:02:50,570 --> 00:02:52,580
We have even video data sets.

40
00:02:52,580 --> 00:02:54,800
So feel free to check that out.

41
00:02:55,010 --> 00:02:56,630
Um that's image segmentation.

42
00:02:56,630 --> 00:02:57,530
So there we go.

43
00:02:57,530 --> 00:03:00,740
And what we see here is the formatting of our data set.

44
00:03:00,740 --> 00:03:08,480
And it happens that the way our data sets are formatted suits this um um, 51 um, way of formatting.

45
00:03:08,480 --> 00:03:11,210
So here we have the data set directory.

46
00:03:11,210 --> 00:03:14,870
You have one um directory for the data.

47
00:03:14,870 --> 00:03:16,070
That's for the images.

48
00:03:16,070 --> 00:03:19,970
And then we have another directory which contains the labels exactly what we expect.

49
00:03:19,970 --> 00:03:24,320
Here we have file name one with the extension found and one with the extension.

50
00:03:24,320 --> 00:03:25,580
So that's it.

51
00:03:25,580 --> 00:03:31,190
And then scrolling down you see how to use this in Python or um the command line.

52
00:03:31,400 --> 00:03:35,000
So given that we're working with Python we could simply copy this out.

53
00:03:35,000 --> 00:03:36,950
So let's copy here.

54
00:03:37,220 --> 00:03:41,150
Let's copy this and then get back to our notebook.

55
00:03:41,330 --> 00:03:42,080
There we go.

56
00:03:42,080 --> 00:03:43,490
It's already installed.

57
00:03:43,520 --> 00:03:47,720
Um, we have um, imported our 51.

58
00:03:47,720 --> 00:03:51,020
And then we paste it in our data set directory.

59
00:03:51,350 --> 00:03:55,310
Let's paste it in here we have data set.

60
00:03:55,550 --> 00:03:56,210
Oops.

61
00:03:56,210 --> 00:03:58,550
Let's get back paste that we have data set.

62
00:03:58,550 --> 00:04:04,370
And then we have um you see we have 51 data set from directory.

63
00:04:04,370 --> 00:04:07,010
And we specify data set directory which we've just done.

64
00:04:07,010 --> 00:04:11,690
Then the data set type you see image segmentation directory.

65
00:04:11,690 --> 00:04:13,850
And then we have the name which we've already specified.

66
00:04:13,850 --> 00:04:14,840
So that looks fine.

67
00:04:14,840 --> 00:04:17,180
Now let's run this and see what we get.

68
00:04:17,300 --> 00:04:23,420
We are getting an error data directory um contents data set data does not exist.

69
00:04:23,420 --> 00:04:27,530
Remember from the documentation we had seen okay.

70
00:04:27,530 --> 00:04:30,230
From here we had seen that um.

71
00:04:30,650 --> 00:04:31,910
Oh let's get back up.

72
00:04:31,910 --> 00:04:38,150
We had seen that this if we specify the data set directory then it's going to search for the data directory

73
00:04:38,300 --> 00:04:40,100
uh under the labels directory.

74
00:04:40,100 --> 00:04:47,930
So given that that's not exactly the names we've used here, we are going to follow this uh, second

75
00:04:47,930 --> 00:04:48,500
format.

76
00:04:48,500 --> 00:04:51,410
Here we have the data path and the labels path.

77
00:04:51,410 --> 00:04:53,810
So let's copy this instead.

78
00:04:54,260 --> 00:04:55,850
Um data path labels path.

79
00:04:56,090 --> 00:04:57,200
Uh that should be fine.

80
00:04:57,200 --> 00:04:59,660
So we have let's replace.

81
00:05:00,160 --> 00:05:04,450
Um this data set directory with data path and labels path.

82
00:05:04,450 --> 00:05:07,690
So we'll just go ahead and copy images.

83
00:05:07,690 --> 00:05:12,550
So let's copy that um copy path and then paste in here.

84
00:05:12,820 --> 00:05:15,430
And we do the same for the labels.

85
00:05:15,430 --> 00:05:16,330
So that's it.

86
00:05:16,330 --> 00:05:21,850
We have taken off um this this actually should be with images.

87
00:05:21,850 --> 00:05:34,360
So we should have images and then we oops we paste that in and then we have masks that should caps should

88
00:05:34,360 --> 00:05:35,380
have masks.

89
00:05:35,380 --> 00:05:41,320
Well this is also PNG masks PNG masks okay.

90
00:05:41,320 --> 00:05:41,980
That looks fine.

91
00:05:41,980 --> 00:05:44,440
So now we have data path.

92
00:05:44,920 --> 00:05:48,130
Let's get back a bit to the check out the documentation.

93
00:05:48,130 --> 00:05:50,530
Here we have data path and labels path.

94
00:05:50,530 --> 00:05:53,320
So we just copy this out and paste.

95
00:05:53,500 --> 00:05:55,870
We have data path and labels path.

96
00:05:55,870 --> 00:05:59,230
Let's run this again and see what we obtain.

97
00:05:59,930 --> 00:06:01,700
You still have an error.

98
00:06:01,730 --> 00:06:02,720
Dataset name.

99
00:06:02,810 --> 00:06:05,060
Loading dataset one is not available.

100
00:06:05,060 --> 00:06:11,660
Now, the reason why we're getting this error is because the first time we try to create a dataset with

101
00:06:11,660 --> 00:06:14,600
this dataset with this, um, specific name.

102
00:06:14,600 --> 00:06:17,570
And so this has already been stored in the database.

103
00:06:17,570 --> 00:06:21,200
So we could only load this data set at this point.

104
00:06:21,290 --> 00:06:24,740
So that said we should change this to to run that again.

105
00:06:24,740 --> 00:06:29,390
And we should now be able to get some outputs here.

106
00:06:29,390 --> 00:06:30,740
We get another error.

107
00:06:30,740 --> 00:06:32,750
But you could see where it is that it started loading.

108
00:06:32,750 --> 00:06:38,420
So uh, one out of 923 elements um or samples started loading.

109
00:06:38,420 --> 00:06:44,900
We are getting this next error because the way we named our mass was different from the way we named

110
00:06:44,900 --> 00:06:45,920
the images.

111
00:06:45,920 --> 00:06:50,990
You could see here we have this key error seq and it's number 0001.

112
00:06:50,990 --> 00:06:58,130
So what we have to do now is rename our masks such that they are all um img instead of seq.

113
00:06:58,130 --> 00:07:04,250
Now to rename this, we also have to take into consideration the fact that right from the beginning

114
00:07:04,250 --> 00:07:06,320
we took this to be seq.

115
00:07:06,320 --> 00:07:10,790
So instead of um this we would have img img.

116
00:07:10,910 --> 00:07:18,380
And then here we would have img and then img and then img.

117
00:07:18,410 --> 00:07:19,040
Okay.

118
00:07:19,040 --> 00:07:21,890
So that's what we're going to have now instead of seq.

119
00:07:22,190 --> 00:07:23,150
Um.

120
00:07:23,900 --> 00:07:27,290
Let's go ahead and rename this.

121
00:07:27,290 --> 00:07:30,950
So that should be before or let's let's just put it in here.

122
00:07:30,950 --> 00:07:33,380
So we're going to import OS.

123
00:07:33,410 --> 00:07:34,940
Import OS.

124
00:07:34,940 --> 00:07:36,920
But this looks like it's already been imported.

125
00:07:36,920 --> 00:07:38,630
So we don't need to import this okay.

126
00:07:38,630 --> 00:07:39,590
So we have that.

127
00:07:39,590 --> 00:07:45,080
And then now what we want to do is rename each and every file we have in here.

128
00:07:45,080 --> 00:07:48,800
Now we have our image path in path and annotation path.

129
00:07:48,800 --> 00:07:51,530
At this point we're supposing that we are yet to split.

130
00:07:51,530 --> 00:07:55,400
That is we are yet to um move these files.

131
00:07:55,400 --> 00:07:59,420
So we'll just work on the original folders.

132
00:07:59,420 --> 00:08:08,720
That said, what we want to do at this point is take in the initial path, and then we want to have

133
00:08:08,720 --> 00:08:13,010
the renamed renamed uh, path or renamed file.

134
00:08:13,010 --> 00:08:15,080
So initial file path.

135
00:08:15,080 --> 00:08:18,350
And then we have the initial file path.

136
00:08:18,350 --> 00:08:20,960
And then we have now the renamed file path.

137
00:08:20,960 --> 00:08:24,680
And then we are going to go through all these different images.

138
00:08:24,680 --> 00:08:33,230
So for I or for image in the list let's take the list of all images.

139
00:08:33,440 --> 00:08:34,610
Um in path.

140
00:08:35,420 --> 00:08:38,660
For this list we are going to do the renaming.

141
00:08:38,780 --> 00:08:42,170
Now this image here is going to be the file name.

142
00:08:42,170 --> 00:08:44,840
So let's let's print it out so you could see that clearly.

143
00:08:44,870 --> 00:08:49,850
Print the image then break out of this loop.

144
00:08:50,120 --> 00:08:51,110
There we go.

145
00:08:51,110 --> 00:08:54,050
Let's take this off and print it out.

146
00:08:54,050 --> 00:08:59,690
So once that's printed we see we have the the file name right here.

147
00:08:59,690 --> 00:09:08,900
So what we want to do now is take rename and then take the image path the image path plus this image

148
00:09:08,930 --> 00:09:11,090
or the plus file name.

149
00:09:11,300 --> 00:09:13,100
Or let's let's just call this file name.

150
00:09:13,100 --> 00:09:17,510
So plus a file name we want to have the file name.

151
00:09:17,510 --> 00:09:20,810
And then let's make it shorter file name.

152
00:09:20,810 --> 00:09:21,980
So there we go.

153
00:09:21,980 --> 00:09:25,400
We have file name f name okay.

154
00:09:25,400 --> 00:09:27,410
So we have the file name.

155
00:09:27,410 --> 00:09:29,780
And then we add this to the image path.

156
00:09:29,780 --> 00:09:34,640
And then once we add this to the image path we have this initial file path.

157
00:09:34,640 --> 00:09:40,490
Now for the final or the renamed file path we want to have the same image path.

158
00:09:40,490 --> 00:09:45,350
But plus this time around um image.

159
00:09:45,860 --> 00:09:52,700
And then that's we have the image and then um, plus the file name.

160
00:09:52,700 --> 00:09:55,070
But starting from the, the underscore.

161
00:09:55,070 --> 00:09:58,310
So we have zero one, two three starting from three.

162
00:09:58,310 --> 00:10:02,960
So we want to have um f name starting from three.

163
00:10:02,960 --> 00:10:03,500
That's it.

164
00:10:03,500 --> 00:10:07,910
Now you should note that we don't need to do this for the images because they already have IMG.

165
00:10:07,940 --> 00:10:10,130
We just need to do this for the annotation.

166
00:10:10,130 --> 00:10:12,950
So take this off and do annotation okay.

167
00:10:12,950 --> 00:10:13,700
So that's it.

168
00:10:13,700 --> 00:10:16,040
So this should work fine.

169
00:10:16,040 --> 00:10:18,350
Let's run this and see what we obtain.

170
00:10:18,350 --> 00:10:23,810
Then we just refresh uh no such file or directory.

171
00:10:23,810 --> 00:10:28,400
The reason why we're getting this error is because this was supposed to be annotation path and not the

172
00:10:28,400 --> 00:10:29,120
image path.

173
00:10:29,120 --> 00:10:31,490
So we have annotation run that again.

174
00:10:31,490 --> 00:10:37,220
And all our files now should be um, modified or their names should be modified.

175
00:10:37,220 --> 00:10:43,370
So if we open up masks you should see that or still have sex.

176
00:10:43,370 --> 00:10:45,410
Let's get back and refresh.

177
00:10:46,010 --> 00:10:51,170
Um, then data set PNG masks, open up masks.

178
00:10:51,170 --> 00:10:57,740
And normally they should be fine because we didn't run into any error when running this here.

179
00:10:57,740 --> 00:10:58,670
Okay, so that's fine.

180
00:10:58,670 --> 00:11:00,500
So now we've renamed our files.

181
00:11:00,500 --> 00:11:06,590
Now what we'll do is that we are just going to, um, disconnect and then restart so that we make sure

182
00:11:06,590 --> 00:11:13,130
that both the training and validation sets now, um, have this new, um, file names.

183
00:11:13,130 --> 00:11:16,670
We've restarted the notebook and running this cell.

184
00:11:16,670 --> 00:11:21,770
Now we should get our data set loaded or we still have this problem.

185
00:11:21,770 --> 00:11:25,970
So to run that again and then they should be fine.

186
00:11:25,970 --> 00:11:32,750
You see we have our 923 different samples and we have the sample fields described.

187
00:11:32,750 --> 00:11:34,160
First of all we have the name.

188
00:11:34,160 --> 00:11:41,180
We have the media type image, um, number of samples, uh, whereas the data is persistent or not.

189
00:11:41,390 --> 00:11:47,570
Um tags, no tags, sample fields, ID file, path tags, metadata.

190
00:11:47,660 --> 00:11:49,040
Um ground truth.

191
00:11:49,040 --> 00:11:56,240
Now for this specific sample or for or picking one sample, we could find that it has an ID.

192
00:11:56,240 --> 00:12:03,530
So um, 51 gives the sample each sample an ID uh the major type is specified the file path.

193
00:12:03,530 --> 00:12:05,210
So in this case is this image.

194
00:12:05,510 --> 00:12:10,010
And then um no tag, um, no metadata.

195
00:12:10,040 --> 00:12:11,510
Then the ground truth.

196
00:12:11,510 --> 00:12:15,380
Now this ground truth here is simply our label because we have the input.

197
00:12:15,380 --> 00:12:16,190
That's the image.

198
00:12:16,190 --> 00:12:17,480
And then we have the label.

199
00:12:17,480 --> 00:12:20,510
Now it's segmentation um label.

200
00:12:20,510 --> 00:12:22,250
So that's why this is specified.

201
00:12:22,250 --> 00:12:23,330
And then we.

202
00:12:23,480 --> 00:12:27,080
Have the ID then we have the mask path.

203
00:12:27,080 --> 00:12:28,820
So here we have the image path.

204
00:12:28,820 --> 00:12:31,160
And here we have the mask path.

205
00:12:31,160 --> 00:12:34,460
So this is what we obtain when we print out the data set.

206
00:12:34,460 --> 00:12:35,570
That's a summary.

207
00:12:35,570 --> 00:12:42,050
And when we print out the data set here that's a few samples from our data set.

208
00:12:42,050 --> 00:12:44,630
Now at this point we could go ahead and launch the app.

209
00:12:44,630 --> 00:12:52,970
So we could do um session, session um, 51 launch app, launch app.

210
00:12:52,970 --> 00:12:55,220
And then we specify simply data set.

211
00:12:55,220 --> 00:12:57,620
We now have our app launched.

212
00:12:57,650 --> 00:13:04,130
Don't forget to give the project a star on GitHub, and also join the 51 slack community to meet other,

213
00:13:04,340 --> 00:13:08,000
um, 51 users and computer vision enthusiasts.

214
00:13:08,360 --> 00:13:10,160
Um, okay, so that's it.

215
00:13:10,160 --> 00:13:14,870
So you see, it takes a little bit, it takes a little while and then comes up.

216
00:13:14,870 --> 00:13:16,130
So there we go.

217
00:13:16,130 --> 00:13:22,070
You could see already um, this interface let's reduce so you could see this much clearer.

218
00:13:22,400 --> 00:13:23,540
There we go.

219
00:13:23,540 --> 00:13:24,020
Okay.

220
00:13:24,020 --> 00:13:25,880
So we have this interface.

221
00:13:25,880 --> 00:13:28,310
Well it's taking a little bit too much time.

222
00:13:28,310 --> 00:13:29,660
Let me refresh this.

223
00:13:29,660 --> 00:13:30,950
Let me run this again.

224
00:13:31,430 --> 00:13:35,030
Um, and it should be coming up soon.

225
00:13:35,690 --> 00:13:39,260
This now shows up and you could see at the top here.

226
00:13:39,290 --> 00:13:40,190
Clothing data set.

227
00:13:40,190 --> 00:13:44,660
One could select another data set in case we had created another data set.

228
00:13:44,660 --> 00:13:47,090
Then we have the filters.

229
00:13:47,090 --> 00:13:50,330
We have the meta data uh which you could open up.

230
00:13:50,330 --> 00:13:55,130
Like you could see this with um, metadata with the see there's no result.

231
00:13:55,130 --> 00:13:56,780
There's no height yet.

232
00:13:56,780 --> 00:13:59,180
But we could actually put in this information.

233
00:13:59,180 --> 00:14:06,530
Then looking at the levels we have the ground truth and we have ID max path, um, tags.

234
00:14:06,530 --> 00:14:08,570
And then we have primitives.

235
00:14:08,570 --> 00:14:10,130
That's ID and file path.

236
00:14:10,130 --> 00:14:14,150
Looking to the right we can see our annotated images.

237
00:14:14,150 --> 00:14:21,080
And instead of getting to open up our view images like this in some folder, and then having to get

238
00:14:21,080 --> 00:14:23,090
back and check out the masks like this.

239
00:14:23,090 --> 00:14:31,190
For example, you have this tool combined together for each and every image and also colored for per

240
00:14:31,190 --> 00:14:32,090
per class.

241
00:14:32,090 --> 00:14:34,430
So each class has a specific color.

242
00:14:34,430 --> 00:14:37,880
Now if you to open any one of this you just need to click on it.

243
00:14:37,880 --> 00:14:39,740
And this one loads up.

244
00:14:39,740 --> 00:14:42,440
Once it's loaded you could have here the primitives.

245
00:14:42,440 --> 00:14:44,330
There's the ID and the file path.

246
00:14:44,330 --> 00:14:50,360
Then you could also check out the metadata in case you have given 51 this information.

247
00:14:50,360 --> 00:14:54,680
So um, hovering on the image like we start from those lenses.

248
00:14:54,680 --> 00:15:01,910
You see, uh, you could read out there clearly 47 that's to show, to tell us that the sunglasses have

249
00:15:01,910 --> 00:15:04,700
ID 47 in our, um, labels.

250
00:15:04,700 --> 00:15:07,460
And then you have the max path and the, the ID.

251
00:15:07,460 --> 00:15:10,970
So as you change this, you see, it goes now to 55.

252
00:15:11,210 --> 00:15:15,860
Um, next you could check out shows, for example, 39 socks.

253
00:15:16,100 --> 00:15:17,690
Uh, 45.

254
00:15:17,840 --> 00:15:19,040
Um, and that's it.

255
00:15:19,040 --> 00:15:24,890
You could check out here, 19 you could check out this, uh, pullover five and you could check out

256
00:15:24,890 --> 00:15:25,430
skin.

257
00:15:25,430 --> 00:15:26,990
That's 41.

258
00:15:26,990 --> 00:15:29,720
Now let's get out of this view.

259
00:15:29,720 --> 00:15:33,110
We could, uh, pick another person.

260
00:15:33,110 --> 00:15:34,610
So let's pick a male.

261
00:15:34,940 --> 00:15:41,030
Um, here we have something similar to what we had already here.

262
00:15:41,030 --> 00:15:50,360
Again, we have those different levels from hair, skin, um, shirt, coat, um, trouser shoe or maybe

263
00:15:50,360 --> 00:15:50,900
socks.

264
00:15:50,900 --> 00:15:51,830
So that's it.

265
00:15:51,830 --> 00:15:55,100
Now you could check take out the ground truth by simply clicking on this.

266
00:15:55,100 --> 00:15:59,960
So once I click on our ground truth, um, I have the original image.

267
00:15:59,960 --> 00:16:06,770
And when I click back at the ground truth, you see, I have now the original image with its mask.

268
00:16:06,770 --> 00:16:09,560
We could also swipe and move to the next image.

269
00:16:09,560 --> 00:16:11,060
So we load the next image.

270
00:16:11,060 --> 00:16:13,460
Here you have the color settings.

271
00:16:13,730 --> 00:16:14,420
Click on this.

272
00:16:14,420 --> 00:16:15,950
You have color settings.

273
00:16:16,070 --> 00:16:18,560
I could modify this opacity of the levels.

274
00:16:18,560 --> 00:16:20,720
So let's reduce this opacity.

275
00:16:21,050 --> 00:16:24,860
Um and then we have save as default.

276
00:16:24,860 --> 00:16:28,520
You see that reducing the opacity makes it look much more transparent.

277
00:16:28,520 --> 00:16:33,710
Let's increase that and then um save as default.

278
00:16:33,710 --> 00:16:38,120
And you see that it looks it uh, it kind of now covers the, the image.

279
00:16:38,120 --> 00:16:42,920
So depending on how you want to view your data set you could modify that opacity.

280
00:16:42,950 --> 00:16:45,110
Now let's get back here.

281
00:16:45,290 --> 00:16:47,180
We'll start by filtering by mask path.

282
00:16:47,210 --> 00:16:51,680
We'll skip the filter by ID and dive straight into filter by mask path.

283
00:16:51,680 --> 00:16:55,280
So depending on the path you want to filter the the image.

284
00:16:55,280 --> 00:16:58,700
So let's say we have 0001.

285
00:16:58,700 --> 00:17:01,520
Let's see you see it filters that image.

286
00:17:01,520 --> 00:17:05,840
And we have this image um 001.

287
00:17:06,020 --> 00:17:11,780
You could see by opening up opening up this um, image here that this is exactly what we expect.

288
00:17:11,780 --> 00:17:16,070
So this is the image 00001.

289
00:17:16,430 --> 00:17:22,760
Um, this is this image with the size, the type file size, and, um, the folder given.

290
00:17:22,760 --> 00:17:27,500
Then when we get back here, you could see that same exact same image.

291
00:17:27,500 --> 00:17:32,180
So you could simply or if you want to see a specific image you could simply filter.

292
00:17:32,180 --> 00:17:37,940
So let's say zero 0 or 0 0990099.

293
00:17:38,240 --> 00:17:39,560
Uh no results.

294
00:17:39,740 --> 00:17:43,880
Um, most probably this has been sent to the validation data set.

295
00:17:43,880 --> 00:17:45,200
So that's why we're not seeing that.

296
00:17:45,410 --> 00:17:51,980
Well we made an error and instead searched in this ID so let's check that out instead in the mask.

297
00:17:51,980 --> 00:18:00,590
So we have 0099 the mask path we should obtain the that that specific image.

298
00:18:00,950 --> 00:18:02,090
And there we go.

299
00:18:02,090 --> 00:18:03,170
We obtain our image.

300
00:18:03,170 --> 00:18:08,270
So this is that 0099 um dot png file.

301
00:18:08,270 --> 00:18:14,720
Now another interesting thing you could do is actually to search again for another let's say six.

302
00:18:14,810 --> 00:18:17,930
You see now we have this, we have selected this two.

303
00:18:17,930 --> 00:18:22,850
So our view right here will be made of only these two images.

304
00:18:22,850 --> 00:18:25,760
So you may have some specific criteria.

305
00:18:25,760 --> 00:18:32,600
You or based on some criteria you want to filter uh, some parts of the data set so that you can just

306
00:18:32,600 --> 00:18:34,370
view only those.

307
00:18:34,370 --> 00:18:34,880
And.

308
00:18:35,280 --> 00:18:42,600
Um, the 51 app makes it very easy to carry out these kinds of filtering and visualizations.

309
00:18:42,600 --> 00:18:46,530
Then from here, another thing you could do is simply save this view.

310
00:18:46,530 --> 00:18:50,610
So if you're interested in only these two images, then you could save that view.

311
00:18:50,610 --> 00:18:54,810
So let me say um save current filters as view.

312
00:18:54,810 --> 00:19:01,110
So your view name we'll call it zero 99 um 1000.

313
00:19:01,110 --> 00:19:04,170
So zero zero 99 okay.

314
00:19:04,380 --> 00:19:08,880
Uh, just some random, um, images.

315
00:19:08,880 --> 00:19:10,110
There we go.

316
00:19:10,110 --> 00:19:10,740
Color.

317
00:19:10,740 --> 00:19:14,010
Let's let's maintain that gray and then we save the view.

318
00:19:14,250 --> 00:19:15,240
So that's it.

319
00:19:15,630 --> 00:19:17,880
So we have our views now saved.

320
00:19:18,000 --> 00:19:24,660
Um, we should get back to the we could get back to the original, um, view that's without having to

321
00:19:24,660 --> 00:19:25,260
filter.

322
00:19:25,260 --> 00:19:27,390
So let's take this off.

323
00:19:29,400 --> 00:19:30,480
Or let's reset this.

324
00:19:30,480 --> 00:19:32,910
Actually resetting this will take all this off.

325
00:19:32,910 --> 00:19:36,720
So now we have the original view with our 923 samples.

326
00:19:36,720 --> 00:19:42,180
And then each time we want to view this, you see each time I want to view this I just need to click.

327
00:19:42,180 --> 00:19:43,350
And there we go.

328
00:19:43,350 --> 00:19:44,160
Just click on that.

329
00:19:44,160 --> 00:19:47,610
And uh, my two images, um show up.

330
00:19:47,610 --> 00:19:50,130
That's it for this section on data preparation.

331
00:19:50,130 --> 00:19:56,550
The next section we'll dive into modeling and training with a sick farmer model, which is available

332
00:19:56,550 --> 00:19:57,660
on Hugging Face.