1
00:00:02,000 --> 00:00:08,000
Now we will train the Yolo v eight segmentation model on the pothole images dataset that we have created

2
00:00:08,000 --> 00:00:10,000
in the previous lecture.

3
00:00:10,000 --> 00:00:16,000
So in this video tutorial we will see how we can train the YOLO eight segmentation model on the porthole

4
00:00:16,000 --> 00:00:20,000
images dataset that we have created in the previous lecture.

5
00:00:20,000 --> 00:00:26,000
I have already prepared the Notebook Google CoLab notebook For this lecture I will guide you as complete

6
00:00:26,000 --> 00:00:33,000
step by step of the complete overflow that how you can train your YOLO v eight segmentation model on

7
00:00:33,000 --> 00:00:34,000
the pothole dataset.

8
00:00:34,000 --> 00:00:40,000
If you follow this complete video tutorial, you will be able to train the YOLO segmentation model on

9
00:00:40,000 --> 00:00:43,000
any of the custom dataset.

10
00:00:43,000 --> 00:00:44,000
So let's get started.

11
00:00:44,000 --> 00:00:49,000
Before running the script, please make sure that you have selected the runtime as GPU.

12
00:00:49,000 --> 00:00:53,000
So my hardware accelerator is being set as GPU, that's why.

13
00:00:53,000 --> 00:00:53,000
Okay.

14
00:00:53,000 --> 00:00:57,000
So in the first step I'm importing all the required libraries.

15
00:00:57,000 --> 00:01:04,000
I only need the library image library to display the input or output image into the Google CoLab notebook.

16
00:01:04,000 --> 00:01:10,000
So we need the image library to display any input or output image in the Google CoLab notebook.

17
00:01:10,000 --> 00:01:12,000
Okay, so just run this cell.

18
00:01:14,000 --> 00:01:14,000
For next.

19
00:01:14,000 --> 00:01:20,000
I am just importing cloning the GitHub repo of YOLO V8 Ultralytics over here.

20
00:01:20,000 --> 00:01:25,000
Okay, so just cloning the YOLO v8 ultralytics GitHub repo over here.

21
00:01:25,000 --> 00:01:26,000
So it's done.

22
00:01:26,000 --> 00:01:28,000
Now you can see that over here.

23
00:01:28,000 --> 00:01:30,000
So here we have the Ultralytics.

24
00:01:30,000 --> 00:01:37,000
If you go over here, we have the YOLO V8 over here and then we can see the detection and segmentation

25
00:01:37,000 --> 00:01:37,000
folder.

26
00:01:37,000 --> 00:01:38,000
So that's it.

27
00:01:38,000 --> 00:01:45,000
So now in the next step, what we need to set, we need to set this cloned folder which we have cloned

28
00:01:45,000 --> 00:01:46,000
over here as our current directory.

29
00:01:46,000 --> 00:01:51,000
So just copy path and just paste this path over here and just run this.

30
00:01:51,000 --> 00:01:58,000
Now in the next step before running the training prediction or validation script, we need to set install

31
00:01:58,000 --> 00:02:04,000
all the required libraries that are necessary to run this YOLO V8 or train the YOLO V8 segmentation

32
00:02:04,000 --> 00:02:05,000
model.

33
00:02:05,000 --> 00:02:08,000
So or do the prediction in the further case as well.

34
00:02:08,000 --> 00:02:13,000
So here using this we will install all the required libraries or the dependencies that are necessary

35
00:02:13,000 --> 00:02:17,000
to run the training prediction or the validation script.

36
00:02:17,000 --> 00:02:22,000
So if you skip this step while running the training validation or the prediction script, you will definitely

37
00:02:22,000 --> 00:02:27,000
face the error that the Hydra library is not installed because you haven't installed the Hydra library.

38
00:02:27,000 --> 00:02:30,000
Then you also face the error that PI test some libraries not installed.

39
00:02:30,000 --> 00:02:33,000
So you can see that some some libraries.

40
00:02:33,000 --> 00:02:39,000
This is appearing requirements already satisfied because some libraries are by default installed in

41
00:02:39,000 --> 00:02:40,000
the Google CoLab.

42
00:02:40,000 --> 00:02:42,000
We don't need to install those libraries.

43
00:02:42,000 --> 00:02:42,000
Okay.

44
00:02:42,000 --> 00:02:45,000
While some libraries are not installed.

45
00:02:45,000 --> 00:02:47,000
So you can see that these are being downloaded over here.

46
00:02:47,000 --> 00:02:54,000
So this step is merged because if you don't run this step or skip this step, you will definitely face

47
00:02:54,000 --> 00:02:54,000
the error that are forming.

48
00:02:54,000 --> 00:02:59,000
The emperor is not installed, Hydra or Matplotlib or Seaborn libraries not installed.

49
00:02:59,000 --> 00:03:03,000
So the only dependencies or the required libraries are being installed next.

50
00:03:03,000 --> 00:03:08,000
As we are doing segmentation over here, then we will need to go to YOLO V8.

51
00:03:09,000 --> 00:03:10,000
And the segmentation folder.

52
00:03:10,000 --> 00:03:14,000
We need to go to this folder and check this folder as our current directory.

53
00:03:14,000 --> 00:03:17,000
I've just have just set this folder as my current directory.

54
00:03:17,000 --> 00:03:21,000
Then we need to see what is our present working directory.

55
00:03:21,000 --> 00:03:25,000
So the segmentation is being set as my present working directory.

56
00:03:25,000 --> 00:03:30,000
So now in the next step what we need to do is we need to export this data set from Roboflow.

57
00:03:30,000 --> 00:03:32,000
Like just copy this from here.

58
00:03:32,000 --> 00:03:33,000
So how did we do this?

59
00:03:33,000 --> 00:03:41,000
Click on export and like the format and just copy this from here and just paste this over in your Google

60
00:03:41,000 --> 00:03:42,000
CoLab notebook.

61
00:03:42,000 --> 00:03:47,000
Okay, So let me just open the Google CoLab notebook and just paste this over here.

62
00:03:47,000 --> 00:03:53,000
So then in this way you can export the data set from Roboflow into your Google CoLab notebook.

63
00:03:53,000 --> 00:03:57,000
So I'm just exporting the data set from Roboflow into the Google CoLab notebook.

64
00:03:57,000 --> 00:03:59,000
So this might take few seconds.

65
00:04:03,000 --> 00:04:04,000
Okay.

66
00:04:04,000 --> 00:04:08,000
So that's I think the dataset will be downloaded.

67
00:04:08,000 --> 00:04:09,000
Okay.

68
00:04:09,000 --> 00:04:11,000
So the dataset is started downloading.

69
00:04:11,000 --> 00:04:11,000
Okay.

70
00:04:12,000 --> 00:04:16,000
So it will be downloaded in the zip format and it will automatically unzip.

71
00:04:16,000 --> 00:04:16,000
Okay.

72
00:04:16,000 --> 00:04:21,000
So now laticeps is consist of only 96 images, so it is easily downloaded.

73
00:04:21,000 --> 00:04:27,000
So now if we click over here we can see the potholes detection dataset and we have the training validation

74
00:04:27,000 --> 00:04:28,000
folders over here.

75
00:04:29,000 --> 00:04:37,000
Okay, so let's now uh, okay, so just set this as our current directory copy path and just set this

76
00:04:37,000 --> 00:04:39,000
as our home directory over here.

77
00:04:40,000 --> 00:04:40,000
Okay.

78
00:04:41,000 --> 00:04:42,000
So now let me see.

79
00:04:42,000 --> 00:04:45,000
What if it is my dataset location from here?

80
00:04:46,000 --> 00:04:50,000
Okay, so here we have the dataset, potholes detection dataset over here.

81
00:04:50,000 --> 00:04:55,000
Okay, let's train our model and see if there is an issue so we can fix it.

82
00:04:55,000 --> 00:05:00,000
And then we will, uh, after the training will take quite some time.

83
00:05:00,000 --> 00:05:04,000
So I'm training the YOLO segmentation model on 100 epochs.

84
00:05:04,000 --> 00:05:04,000
Okay.

85
00:05:04,000 --> 00:05:06,000
What we see over here.

86
00:05:06,000 --> 00:05:08,000
Okay, so there might be an update in the YOLO.

87
00:05:09,000 --> 00:05:11,000
So let me just fix it.

88
00:05:11,000 --> 00:05:12,000
Okay, So.

89
00:05:12,000 --> 00:05:16,000
So if you face ever any issue like this, so you just need to go to.

90
00:05:18,000 --> 00:05:21,000
Like just like your lovin it.

91
00:05:21,000 --> 00:05:22,000
I'll try it.

92
00:05:22,000 --> 00:05:24,000
So just write this over here.

93
00:05:24,000 --> 00:05:25,000
Okay.

94
00:05:25,000 --> 00:05:28,000
And just click on the first link over here.

95
00:05:29,000 --> 00:05:32,000
And just go to ultralytics.

96
00:05:32,000 --> 00:05:33,000
YOLO.

97
00:05:33,000 --> 00:05:36,000
There might be an update, so you might face this error.

98
00:05:36,000 --> 00:05:41,000
But if you work like like this like I'm doing, you will not face this error.

99
00:05:41,000 --> 00:05:48,000
So just go over here and just copy this because here I'm just pasting this error copy and just paste

100
00:05:48,000 --> 00:05:49,000
this over here.

101
00:05:49,000 --> 00:05:54,000
And I hope now our issue will be solved and will not be facing any such issue.

102
00:05:54,000 --> 00:05:56,000
So let's get started.

103
00:05:56,000 --> 00:05:59,000
Let's train the YOLO Age segmentation model.

104
00:06:00,000 --> 00:06:02,000
Okay, so the training is about to start.

105
00:06:02,000 --> 00:06:04,000
Let me see if there is any issue so we can fix it.

106
00:06:04,000 --> 00:06:07,000
If it's working fine that then that would be great.

107
00:06:08,000 --> 00:06:09,000
Okay.

108
00:06:09,000 --> 00:06:15,000
So first of all, it is downloading downloading the Pre-trained model from the GitHub repo into this

109
00:06:15,000 --> 00:06:16,000
Google CoLab notebook.

110
00:06:16,000 --> 00:06:17,000
So.

111
00:06:19,000 --> 00:06:22,000
And now you can see that the Pre-trained model is being downloaded.

112
00:06:23,000 --> 00:06:28,000
We will fine tune this segmentation model on the potholes images.

113
00:06:28,000 --> 00:06:32,000
Okay, so we are just fine tuning this segmentation model on the potholes images.

114
00:06:32,000 --> 00:06:34,000
So that training will take around one hour.

115
00:06:34,000 --> 00:06:39,000
So as the training starts, I will pause the video and when the training complete, I will be back and

116
00:06:39,000 --> 00:06:41,000
then we will discuss the results.

117
00:06:41,000 --> 00:06:42,000
So the training is start.

118
00:06:42,000 --> 00:06:47,000
I am training the YOLO segmentation model on 1.2 epochs, so the training will take time.

119
00:06:47,000 --> 00:06:52,000
So I'm just pausing the video and when the training will complete I will be back and then we will discuss

120
00:06:52,000 --> 00:06:53,000
the results.

121
00:06:57,000 --> 00:07:01,000
Well, we so we have trained the eight segmentation model.

122
00:07:01,000 --> 00:07:08,000
Or you can say that we have fine tuned the YOLO eight segmentation model on potholes images, and we

123
00:07:08,000 --> 00:07:16,000
have trained the segmentation model on 120 epochs with default image size as 640 and here we pass the

124
00:07:16,000 --> 00:07:23,000
data dot Yml file path data dot Yml file contains the train test and validation images file paths and

125
00:07:23,000 --> 00:07:28,000
the model we use to fine tune the model we use to fine tune this.

126
00:07:28,000 --> 00:07:29,000
YOLO.

127
00:07:29,000 --> 00:07:35,000
Yolo v eight comes with five more different models Yolo v eight and is the smallest, but it is less

128
00:07:35,000 --> 00:07:41,000
accurate and it is fast, while YOLO eight X is more accurate, but it is less fast as compared to the

129
00:07:41,000 --> 00:07:43,000
other YOLO V8 model.

130
00:07:43,000 --> 00:07:50,000
So we have considered a medium level of V8 model for to fine tune our YOLO V8 segmentation model on

131
00:07:50,000 --> 00:07:52,000
the potholes dataset.

132
00:07:52,000 --> 00:07:53,000
Okay.

133
00:07:53,000 --> 00:07:59,000
Or you can say that we are training the V8 segmentation model on the potholes images dataset.

134
00:07:59,000 --> 00:08:06,000
So we have trained the YOLO segmentation model on 120 epochs so we can see that after every epochs,

135
00:08:06,000 --> 00:08:12,000
the mean average precision with 50 and mean average precision with 50 to 95 is improving.

136
00:08:12,000 --> 00:08:19,000
So after training the segmentation model on 120 epochs, here we have the results.

137
00:08:19,000 --> 00:08:20,000
So let me show you the results.

138
00:08:21,000 --> 00:08:26,000
We're here so we can see that the mean average precision with 50.

139
00:08:26,000 --> 00:08:35,000
We got at the end of 120 epochs is 0.353 and mean average precision with 50 to 95 we got is 0.159.

140
00:08:35,000 --> 00:08:37,000
These are not very good.

141
00:08:37,000 --> 00:08:43,000
Mean average precision values are because we have only trained our segmentation model on 96 images.

142
00:08:43,000 --> 00:08:49,000
Our dataset consists of 96 images, but if we train our YOLO segmentation model on a large dataset like

143
00:08:49,000 --> 00:08:56,000
1015 hundred, then we will definitely get a mean average precision above 70%, like 0.73.

144
00:08:56,000 --> 00:09:01,000
Okay, so it's currently 35.3% and it is 15.9%.

145
00:09:01,000 --> 00:09:01,000
Okay.

146
00:09:01,000 --> 00:09:03,000
And our best weights.

147
00:09:03,000 --> 00:09:06,000
So best weights best represents the best weights.

148
00:09:06,000 --> 00:09:11,000
Best weights are the way those weights one are represent the weight on that epoch.

149
00:09:11,000 --> 00:09:15,000
For example, on 118 epoch, our model gives us the best results.

150
00:09:15,000 --> 00:09:17,000
So this will be considered as the best weights.

151
00:09:17,000 --> 00:09:23,000
So best dot represent the best dot, represents the best weights, which means the weights on the epoch,

152
00:09:23,000 --> 00:09:29,000
on which model performs the best, while last.is the weights of the model on the last epoch.

153
00:09:29,000 --> 00:09:32,000
So the last epoch over here is 1/20 epoch.

154
00:09:32,000 --> 00:09:39,000
So the last dot represents the weights of the YOLO train, YOLO segmentation model on the last epoch.

155
00:09:40,000 --> 00:09:42,000
So here we have the model summary.

156
00:09:42,000 --> 00:09:51,000
The mean average precision with 50 we got is 0.532 and the mean average precision with 5295 is 0.218,

157
00:09:51,000 --> 00:09:55,000
which is 21.8% and this is 53.2%.

158
00:09:55,000 --> 00:09:59,000
And our results are saved in run segment train.

159
00:09:59,000 --> 00:10:00,000
So if we see the results.

160
00:10:02,000 --> 00:10:04,000
We stood in this folded framed folder.

161
00:10:04,000 --> 00:10:06,000
Our confusion matrix results.

162
00:10:07,000 --> 00:10:10,000
Training and validation loss results are saved.

163
00:10:10,000 --> 00:10:13,000
So let me check what different files we have in this folder.

164
00:10:13,000 --> 00:10:21,000
So just try checking the files and files in the training folder.

165
00:10:22,000 --> 00:10:25,000
Okay, so check the files in the training folder.

166
00:10:25,000 --> 00:10:31,000
So just click over here and click on LHS and just pass this path from here.

167
00:10:31,000 --> 00:10:38,000
Copy path and like just paste this path over here and let us see what different files we have in this

168
00:10:38,000 --> 00:10:39,000
training folder.

169
00:10:39,000 --> 00:10:44,000
We are all our results are saved, so we have confusion matrix results dot PNG file.

170
00:10:44,000 --> 00:10:46,000
We have the validation predictions as well.

171
00:10:46,000 --> 00:10:51,000
We have the F1 curve, we have the recall curve, we have the precision curves as well.

172
00:10:51,000 --> 00:10:56,000
So let's first validate our custom order on the best weights which we have got over here.

173
00:10:56,000 --> 00:10:56,000
Best Dot.

174
00:10:56,000 --> 00:10:59,000
BT Okay, so we always consider the best split.

175
00:10:59,000 --> 00:11:01,000
So best DOT represents the best period.

176
00:11:01,000 --> 00:11:07,000
So let us evaluate it, the custom order on the best weights so we are validating the custom order on

177
00:11:07,000 --> 00:11:09,000
the validation dataset images.

178
00:11:09,000 --> 00:11:13,000
So if we see over here, our validation dataset images are over here.

179
00:11:13,000 --> 00:11:17,000
So the validation folder contains the validation dataset images.

180
00:11:17,000 --> 00:11:19,000
So these are around 15 images.

181
00:11:19,000 --> 00:11:22,000
So we are validating the Yolo v segment.

182
00:11:22,000 --> 00:11:26,000
Train your the segmentation model on our whole dataset.

183
00:11:26,000 --> 00:11:27,000
Okay.

184
00:11:27,000 --> 00:11:28,000
On these images.

185
00:11:28,000 --> 00:11:30,000
So let's run this.

186
00:11:30,000 --> 00:11:37,000
So we are using validation dot pi script because we are validating our train YOLO segmentation model

187
00:11:37,000 --> 00:11:39,000
on this validation data.

188
00:11:39,000 --> 00:11:44,000
Set images and let's see what mean average precision we get from this.

189
00:11:45,000 --> 00:11:46,000
Okay.

190
00:11:46,000 --> 00:11:54,000
So the mean average precision we got is 53.1% with 50 and mean average precision with 50 to 95 is 22.4%.

191
00:11:54,000 --> 00:11:59,000
Okay, so let's see the model prediction on the validation batch.

192
00:11:59,000 --> 00:12:00,000
So.

193
00:12:00,000 --> 00:12:03,000
So these are the model predictions on the validation batch.

194
00:12:03,000 --> 00:12:05,000
So I'm just displaying this image from here.

195
00:12:05,000 --> 00:12:07,000
Let me show you which image I'm displaying.

196
00:12:08,000 --> 00:12:08,000
Okay.

197
00:12:08,000 --> 00:12:09,000
So just give me a minute.

198
00:12:09,000 --> 00:12:12,000
Just let me just go to the.

199
00:12:13,000 --> 00:12:15,000
Train folder from here.

200
00:12:15,000 --> 00:12:15,000
Okay.

201
00:12:15,000 --> 00:12:16,000
Where is my.

202
00:12:16,000 --> 00:12:17,000
Okay, just.

203
00:12:17,000 --> 00:12:18,000
Just give me a minute.

204
00:12:18,000 --> 00:12:20,000
I'm saying this just.

205
00:12:24,000 --> 00:12:25,000
Or in the training folder.

206
00:12:25,000 --> 00:12:30,000
We have the results for the validation predictions, validation, batch predictions.

207
00:12:30,000 --> 00:12:33,000
So these are the predictions on the validation batch.

208
00:12:33,000 --> 00:12:36,000
So these images are not used for the training.

209
00:12:36,000 --> 00:12:42,000
And so it's always better to have a look and see how our model performs on the validation dataset images

210
00:12:42,000 --> 00:12:46,000
so we can see that the model is able to detect the pothole, do the segmentation as well.

211
00:12:46,000 --> 00:12:50,000
So we have the detection bounding box and plus we are doing the segmentation as well.

212
00:12:50,000 --> 00:12:53,000
We are creating a mask where we have the pothole.

213
00:12:54,000 --> 00:12:55,000
Okay, so that's cool.

214
00:12:56,000 --> 00:12:57,000
Okay.

215
00:12:57,000 --> 00:13:01,000
Let me just I think this is the previous step printed skip results.

216
00:13:01,000 --> 00:13:02,000
So let me just refresh it.

217
00:13:02,000 --> 00:13:03,000
Just give me a minute.

218
00:13:04,000 --> 00:13:04,000
Okay?

219
00:13:04,000 --> 00:13:08,000
So just copy this from here.

220
00:13:08,000 --> 00:13:08,000
Okay?

221
00:13:08,000 --> 00:13:10,000
Let me just go to the.

222
00:13:10,000 --> 00:13:14,000
Okay, just copy this from here and just paste this from here to here.

223
00:13:14,000 --> 00:13:17,000
And let's see, what results do we get from here.

224
00:13:18,000 --> 00:13:21,000
Okay, so this might take some time.

225
00:13:21,000 --> 00:13:26,000
Okay, so these are the images you can see that we are able to detect the pothole.

226
00:13:26,000 --> 00:13:28,000
We have done the segmentation of the pothole as well.

227
00:13:28,000 --> 00:13:32,000
And this is the confidence score and the label that we have the pothole.

228
00:13:33,000 --> 00:13:34,000
Okay.

229
00:13:34,000 --> 00:13:35,000
So here is the confusion matrix.

230
00:13:36,000 --> 00:13:40,000
So confusion matrix basically show that how our model handles different classes.

231
00:13:40,000 --> 00:13:46,000
So so from this confusion matrix, we can see that 50% of our times, our model detected correctly that

232
00:13:46,000 --> 00:13:52,000
there is a pothole, while 50% of the time when there is a pothole, our model was unable to detect

233
00:13:52,000 --> 00:13:53,000
that there is a pothole.

234
00:13:53,000 --> 00:13:55,000
Okay, so.

235
00:13:56,000 --> 00:14:00,000
If you want to see that training and validation losses, the losses.

236
00:14:00,000 --> 00:14:03,000
So just let me just copy path from here.

237
00:14:03,000 --> 00:14:04,000
Okay.

238
00:14:04,000 --> 00:14:10,000
And okay, so let me just copy this from here and let me show you training and validation loss.

239
00:14:10,000 --> 00:14:16,000
So what training and validation losses.

240
00:14:16,000 --> 00:14:20,000
So we will just plot the graph for the training and validation loss.

241
00:14:20,000 --> 00:14:26,000
So just go to over here, just copy path from here and just paste this from here.

242
00:14:26,000 --> 00:14:30,000
So this will now pretend the training and validation losses.

243
00:14:30,000 --> 00:14:34,000
So let's see how our results look like.

244
00:14:34,000 --> 00:14:37,000
So you can see that the loss is continuously decreasing.

245
00:14:37,000 --> 00:14:42,000
So if you train our model on more than 120 epochs, we can say that the loss will further decrease,

246
00:14:42,000 --> 00:14:44,000
while the mean average precision tends to increase.

247
00:14:44,000 --> 00:14:51,000
So if we train our model or increase the images of the dataset like currently we are training the segmentation

248
00:14:52,000 --> 00:14:55,000
on pothole dataset with only 96 images.

249
00:14:55,000 --> 00:15:01,000
So if we increase the dataset up to 500 or 1000, so we will have the better mean average precision

250
00:15:01,000 --> 00:15:02,000
and it will increase.

251
00:15:02,000 --> 00:15:06,000
And if we train our model on higher epochs, the mean average precision will further improve.

252
00:15:06,000 --> 00:15:07,000
Okay.

253
00:15:07,000 --> 00:15:13,000
So in the results dot csv file, we have the model performance after each of the epochs.

254
00:15:13,000 --> 00:15:18,000
Like you can see that we have the model mean average precision after each of the epoch mean average

255
00:15:18,000 --> 00:15:21,000
precision results after each of the epoch.

256
00:15:21,000 --> 00:15:25,000
So in the results dot csv file contains the training losses.

257
00:15:25,000 --> 00:15:29,000
Validation losses mean average precision values after each of the epoch.

258
00:15:29,000 --> 00:15:37,000
Okay, so here I will just download the demo video and see how our model gives us the results on a demo

259
00:15:37,000 --> 00:15:38,000
video when we test it.

260
00:15:38,000 --> 00:15:38,000
Okay.

261
00:15:38,000 --> 00:15:44,000
So I'm just passing the best weights and as I'm doing the prediction, so I'm just passing the predict.py

262
00:15:44,000 --> 00:15:51,000
file and I'm just passing the demo dot before the demo video in which I'm testing my train YOLO segmentation

263
00:15:51,000 --> 00:15:56,000
model or fine tuned YOLO segmentation model on the potholes dataset.

264
00:15:56,000 --> 00:15:56,000
Okay, so.

265
00:15:58,000 --> 00:16:00,000
This might take some time to execute.

266
00:16:00,000 --> 00:16:08,000
Okay, so the demo video is not very large, so I think it will divide the complete video into 324 frames.

267
00:16:08,000 --> 00:16:12,000
And it is doing the it is processing each of the frame one by one.

268
00:16:12,000 --> 00:16:18,000
So currently we can see that out of 324 to 24 frames are processed.

269
00:16:18,000 --> 00:16:18,000
Okay.

270
00:16:18,000 --> 00:16:24,000
So it's, uh, processing each of the frame one by one and in response, detect train four.

271
00:16:24,000 --> 00:16:25,000
Our results are saved.

272
00:16:25,000 --> 00:16:28,000
Okay, so the train four, our results are saved.

273
00:16:28,000 --> 00:16:32,000
Okay, so just let's run this cell and see what results do we get.

274
00:16:32,000 --> 00:16:36,000
Okay, so the output video will be displayed in front over here.

275
00:16:36,000 --> 00:16:42,000
Then I will download this video and show you what results do we get from here.

276
00:16:42,000 --> 00:16:44,000
So this might take few seconds more.

277
00:16:46,000 --> 00:16:50,000
So the next step let the video be displayed in the Google CoLab notebook.

278
00:16:50,000 --> 00:16:55,000
Then I will download this video from here and let me show you the result.

279
00:16:55,000 --> 00:17:02,000
So let me download this video from here and let me just play this video and let me navigate my screen

280
00:17:02,000 --> 00:17:03,000
towards this video.

281
00:17:03,000 --> 00:17:09,000
So now you can see that our model train, your segmentation model, is able to detect successfully the

282
00:17:09,000 --> 00:17:11,000
potholes here as well.

283
00:17:11,000 --> 00:17:14,000
So the results are quite impressive.

284
00:17:14,000 --> 00:17:16,000
So this is all from this video and audio.

285
00:17:16,000 --> 00:17:17,000
See you all in the next video.

286
00:17:17,000 --> 00:17:19,000
Till then, bye bye.