1 00:00:02,000 --> 00:00:03,000 Hello everyone. 2 00:00:03,000 --> 00:00:10,000 In this video tutorial we will see how we can train Yolo V8 on custom dataset of potholes. 3 00:00:10,000 --> 00:00:15,000 So the first step of training YOLO V8 on any custom dataset is to collect the dataset. 4 00:00:15,000 --> 00:00:19,000 So I have found this pothole image dataset on kaggle.com. 5 00:00:19,000 --> 00:00:26,000 So first thing you need to do is to go to Kaggle and create an account if you wish to download any dataset 6 00:00:26,000 --> 00:00:27,000 from Kaggle. 7 00:00:27,000 --> 00:00:29,000 So I have already created my account. 8 00:00:29,000 --> 00:00:34,000 So in the first step I will download this dataset from here. 9 00:00:34,000 --> 00:00:35,000 So I will just click on download. 10 00:00:37,000 --> 00:00:38,000 For the dataset. 11 00:00:38,000 --> 00:00:42,000 It might take few minutes or seconds to get downloaded. 12 00:00:43,000 --> 00:00:44,000 So I will wait. 13 00:00:51,000 --> 00:00:52,000 Well know the dataset. 14 00:00:52,000 --> 00:00:54,000 It can be downloaded. 15 00:00:54,000 --> 00:00:56,000 So will just go to this dataset over here. 16 00:00:57,000 --> 00:00:59,000 And just copy this data set. 17 00:01:00,000 --> 00:01:03,000 And here I have created a folder for this project. 18 00:01:03,000 --> 00:01:09,000 So I will just paste this dataset over here and just unzip it that this dataset. 19 00:01:09,000 --> 00:01:11,000 So I'll click on extract files. 20 00:01:11,000 --> 00:01:15,000 So okay, so it might take few seconds to extract. 21 00:01:15,000 --> 00:01:17,000 Okay, so we have the dataset. 22 00:01:17,000 --> 00:01:21,000 Okay over here, but the dataset is not annotated. 23 00:01:21,000 --> 00:01:24,000 So we first need to annotate this dataset. 24 00:01:24,000 --> 00:01:26,000 So the dataset is quite large. 25 00:01:26,000 --> 00:01:35,000 It consists of around 620 images, so that is fine for this for the training process. 26 00:01:35,000 --> 00:01:41,000 So first of all, to annotate the dataset, I will go to roboflow. 27 00:01:43,000 --> 00:01:43,000 That's right. 28 00:01:44,000 --> 00:01:44,000 Right. 29 00:01:44,000 --> 00:01:45,000 Roboflow on Google. 30 00:01:45,000 --> 00:01:48,000 And just open this first link over here. 31 00:01:51,000 --> 00:01:52,000 And just click on sign up. 32 00:01:58,000 --> 00:01:59,000 Okay. 33 00:01:59,000 --> 00:02:01,000 So in the first step, we need to create an account. 34 00:02:02,000 --> 00:02:04,000 I will sign in with. 35 00:02:04,000 --> 00:02:07,000 So the first step, I will create an account over here. 36 00:02:07,000 --> 00:02:11,000 I don't have account, uh, on this email on roboflow. 37 00:02:11,000 --> 00:02:15,000 So here I'm just creating an account from start. 38 00:02:18,000 --> 00:02:19,000 Arcade. 39 00:02:20,000 --> 00:02:21,000 So let's see. 40 00:02:21,000 --> 00:02:25,000 It might take a few seconds, so let's see and wait and see. 41 00:02:29,000 --> 00:02:30,000 Okay. 42 00:02:33,000 --> 00:02:35,000 And so confirm your account in full. 43 00:02:35,000 --> 00:02:37,000 Accept the terms of service and policy. 44 00:02:37,000 --> 00:02:38,000 Continue. 45 00:02:39,000 --> 00:02:41,000 I would need to follow the same steps. 46 00:02:41,000 --> 00:02:43,000 Please don't skip any of the steps you. 47 00:02:43,000 --> 00:02:45,000 Then you might face an issue. 48 00:02:45,000 --> 00:02:48,000 Okay, so here is my voice. 49 00:02:48,000 --> 00:02:56,000 So what are what you will you be using this workspace board for the academy purpose workspace name and 50 00:02:56,000 --> 00:02:59,000 I can write my name and going sir. 51 00:03:00,000 --> 00:03:00,000 Okay. 52 00:03:02,000 --> 00:03:09,000 And click on continue in one, then skip this step and then for community purpose. 53 00:03:13,000 --> 00:03:15,000 Okay, That's cool. 54 00:03:18,000 --> 00:03:20,000 And then you need to go to the settings. 55 00:03:21,000 --> 00:03:26,000 And then click over here and you need to make sure that your workspace is on the public plan. 56 00:03:26,000 --> 00:03:32,000 So if your workspace is not on the public plan, if it's on some other plan, then you will not be able 57 00:03:32,000 --> 00:03:35,000 to export the dataset or annotate the dataset over here. 58 00:03:35,000 --> 00:03:39,000 So please make sure that your workspace should be on the public plan. 59 00:03:39,000 --> 00:03:44,000 Okay, so now I will go to the robot flow again and create new project. 60 00:03:45,000 --> 00:03:46,000 Okay. 61 00:03:46,000 --> 00:03:47,000 And now. 62 00:03:49,000 --> 00:03:49,000 Okay. 63 00:03:49,000 --> 00:03:55,000 So basically I am performing pothole detection, so I will just go click on object detection. 64 00:03:55,000 --> 00:04:01,000 So if you are performing instance segmentation or semantic segmentation, you can click from here or 65 00:04:01,000 --> 00:04:03,000 choose the required option from here. 66 00:04:03,000 --> 00:04:04,000 So what are you detecting? 67 00:04:04,000 --> 00:04:06,000 I'm detecting potholes. 68 00:04:06,000 --> 00:04:06,000 Okay. 69 00:04:08,000 --> 00:04:09,000 So project name. 70 00:04:10,000 --> 00:04:13,000 Pothole detection. 71 00:04:14,000 --> 00:04:16,000 Yolo v eight. 72 00:04:16,000 --> 00:04:20,000 So this is the name of our project and click on Create Public Project. 73 00:04:22,000 --> 00:04:25,000 Okay, So first of all, we need to select files. 74 00:04:25,000 --> 00:04:26,000 We need to annotate. 75 00:04:26,000 --> 00:04:31,000 So as you know that we have downloaded the dataset from Kaggle, we need to annotate the dataset. 76 00:04:31,000 --> 00:04:33,000 So I will click on select files. 77 00:04:34,000 --> 00:04:37,000 And will go to your gallery. 78 00:04:37,000 --> 00:04:39,000 It was from here. 79 00:04:39,000 --> 00:04:40,000 I will just go to. 80 00:04:41,000 --> 00:04:42,000 Was that us? 81 00:04:42,000 --> 00:04:50,000 So here from here, I'll just click on here and then click on archive and then select all. 82 00:04:50,000 --> 00:04:53,000 And then I will upload these files over here. 83 00:04:54,000 --> 00:04:56,000 So it is processing. 84 00:04:56,000 --> 00:04:59,000 So it will be uploading all the files over here. 85 00:04:59,000 --> 00:05:05,000 So after uploading these files then I will annotate each of the image in that in this dataset. 86 00:05:05,000 --> 00:05:06,000 Okay. 87 00:05:08,000 --> 00:05:11,000 So I'm just pausing the video until this processing completes. 88 00:05:11,000 --> 00:05:12,000 Then I will be back here. 89 00:05:16,000 --> 00:05:18,000 The processing of files is done. 90 00:05:18,000 --> 00:05:26,000 So now I will save and continue so that these images are uploaded into my dataset section over here. 91 00:05:26,000 --> 00:05:27,000 Okay. 92 00:05:27,000 --> 00:05:31,000 But you can currently see that we have 608 images and all are not annotated. 93 00:05:31,000 --> 00:05:36,000 Not a single image is annotated and we will annotate all these images. 94 00:05:36,000 --> 00:05:38,000 So first I will click on Save and Continue. 95 00:05:39,000 --> 00:05:43,000 So that all these files are uploaded. 96 00:05:43,000 --> 00:05:47,000 Then I will show you how you can annotate each image over here. 97 00:05:48,000 --> 00:05:52,000 So it might take two seconds or a few minutes to upload these files. 98 00:05:52,000 --> 00:05:53,000 So will stop. 99 00:05:53,000 --> 00:05:57,000 Pause this video and as these files get uploaded, then I will be back. 100 00:06:01,000 --> 00:06:02,000 Guys. 101 00:06:02,000 --> 00:06:09,000 The files are uploaded so it after the files upload are uploaded it appears there which is for annotation. 102 00:06:09,000 --> 00:06:15,000 So currently, as I told you that the image are not annotated, so we need to annotate each of the single 103 00:06:15,000 --> 00:06:16,000 image. 104 00:06:16,000 --> 00:06:22,000 So as I am only one that is working on this project, so I will assign it to myself. 105 00:06:22,000 --> 00:06:27,000 But if you have other group member or a team member who is working this project with you, then you 106 00:06:27,000 --> 00:06:33,000 can invite that teammate with through his email and he can also help you to annotate these images. 107 00:06:33,000 --> 00:06:38,000 So currently I am the only one working on this project, so I will annotate these on these images. 108 00:06:39,000 --> 00:06:40,000 Okay, so. 109 00:06:42,000 --> 00:06:42,000 Here. 110 00:06:42,000 --> 00:06:48,000 Now you can see the uneducated and those you annotate will appear here annotated in the annotated section. 111 00:06:48,000 --> 00:06:51,000 So I will just click on this first one. 112 00:06:52,000 --> 00:06:52,000 Okay. 113 00:06:52,000 --> 00:06:54,000 So you can see this part over here. 114 00:06:54,000 --> 00:07:01,000 So I will just click over here bounding box tool and just draw this over here. 115 00:07:01,000 --> 00:07:02,000 Okay? 116 00:07:03,000 --> 00:07:03,000 So enter here. 117 00:07:03,000 --> 00:07:04,000 Right. 118 00:07:04,000 --> 00:07:05,000 Pothole. 119 00:07:05,000 --> 00:07:10,000 So basically, our class name will be potholes and save. 120 00:07:10,000 --> 00:07:14,000 Okay, so I can also see a pothole over here as well. 121 00:07:14,000 --> 00:07:18,000 So I will draw this bounding drops around here and save this as well. 122 00:07:18,000 --> 00:07:22,000 Okay, so I don't see any other pothole in this complete image. 123 00:07:22,000 --> 00:07:24,000 So let's move towards the next image. 124 00:07:28,000 --> 00:07:34,000 Well, I can see a big pothole over here, so I will just draw a bounding box over here as well. 125 00:07:34,000 --> 00:07:35,000 You can see that. 126 00:07:35,000 --> 00:07:35,000 Okay. 127 00:07:35,000 --> 00:07:38,000 And then click on Save and Enter. 128 00:07:38,000 --> 00:07:38,000 Okay. 129 00:07:38,000 --> 00:07:43,000 So you can just see over here that we have that drawn the bounding box around this pothole and labeled 130 00:07:43,000 --> 00:07:44,000 it as well. 131 00:07:44,000 --> 00:07:49,000 So now go to the another image, the third one, as you can see, a big pothole over here. 132 00:07:49,000 --> 00:07:51,000 So just annotate it. 133 00:07:51,000 --> 00:07:53,000 And just see. 134 00:07:53,000 --> 00:07:57,000 Okay, just click on Save and you can see that we have labeled pothole. 135 00:07:57,000 --> 00:07:59,000 Now let's move towards the next image. 136 00:08:00,000 --> 00:08:03,000 So there is also a big pothole, so let's label it. 137 00:08:04,000 --> 00:08:04,000 Okay. 138 00:08:04,000 --> 00:08:09,000 And then click on Save Enter and you can see that we have the bullet potholes. 139 00:08:09,000 --> 00:08:11,000 Then go towards the next image as well. 140 00:08:11,000 --> 00:08:14,000 So we can see a lot of potholes over here. 141 00:08:14,000 --> 00:08:16,000 So let's annotate each of them. 142 00:08:16,000 --> 00:08:17,000 Okay. 143 00:08:17,000 --> 00:08:25,000 And that's click on Save Enter, then go to this one and then click on Save, Enter, and then select 144 00:08:25,000 --> 00:08:28,000 this one and then click on Save Enter. 145 00:08:28,000 --> 00:08:32,000 Then select this one and click on Save Enter. 146 00:08:32,000 --> 00:08:33,000 Then select this one. 147 00:08:33,000 --> 00:08:38,000 Click on Save, Enter and select this one as well and click on Save Enter. 148 00:08:38,000 --> 00:08:44,000 So in this way I will annotate all the single images in in this folder. 149 00:08:44,000 --> 00:08:46,000 So there are total 608 images. 150 00:08:46,000 --> 00:08:49,000 So I will annotate every single image. 151 00:08:49,000 --> 00:08:49,000 Okay. 152 00:08:50,000 --> 00:08:57,000 So after I annotate all these single images, then I will go to add five images to my dataset. 153 00:08:57,000 --> 00:09:03,000 For example, currently I've annotated five images, so I need to annotate all these 603 images, but 154 00:09:03,000 --> 00:09:06,000 just to show you as you've annotated five images. 155 00:09:06,000 --> 00:09:12,000 So I will click on Add five images to dataset, and here you can select that train valid test plate. 156 00:09:12,000 --> 00:09:18,000 So basically we take 70% for the training, 20% for the validation and 10% for the testing. 157 00:09:18,000 --> 00:09:24,000 So if you click on Add images over here, you can see that it might take few seconds, so please bear 158 00:09:24,000 --> 00:09:25,000 with me. 159 00:09:26,000 --> 00:09:26,000 Pandemic. 160 00:09:26,000 --> 00:09:30,000 So you can see that our data set now has five images. 161 00:09:30,000 --> 00:09:30,000 Okay. 162 00:09:31,000 --> 00:09:33,000 Uh, currently three in the training. 163 00:09:34,000 --> 00:09:35,000 And. 164 00:09:39,000 --> 00:09:40,000 Okay. 165 00:09:40,000 --> 00:09:45,000 So three in the training, one in the validation and one in the test. 166 00:09:45,000 --> 00:09:49,000 So our dataset total has five images one, two, three, four, five. 167 00:09:49,000 --> 00:09:58,000 So in the same way I will annotate this remaining 603 images and then I will also add them into my dataset. 168 00:09:58,000 --> 00:10:01,000 So let's annotate on these images. 169 00:10:01,000 --> 00:10:02,000 Okay? 170 00:10:03,000 --> 00:10:04,000 That's one thing I want to tell you. 171 00:10:04,000 --> 00:10:09,000 For example, if you're if you don't have the image dataset, you have some YouTube video. 172 00:10:10,000 --> 00:10:10,000 Okay. 173 00:10:10,000 --> 00:10:15,000 So how you can separate a image frame by frame from the YouTube video. 174 00:10:15,000 --> 00:10:16,000 Let me explain this to you as well. 175 00:10:20,000 --> 00:10:26,000 Well, I have found this image of a video of potholes on YouTube, but just by writing pothole roads 176 00:10:26,000 --> 00:10:30,000 and I just found this video in the top search. 177 00:10:30,000 --> 00:10:33,000 So just open this video so you can see that in this video. 178 00:10:33,000 --> 00:10:37,000 We have also have potholes like in this video. 179 00:10:37,000 --> 00:10:43,000 So if we want to if we have a less dense site or insufficient dataset so we can collect a dataset from 180 00:10:43,000 --> 00:10:44,000 videos as well. 181 00:10:44,000 --> 00:10:51,000 So you can just need to copy this from here and just add this link over here and just go to over here. 182 00:10:51,000 --> 00:10:56,000 So please wait while we, while we process your video. 183 00:10:56,000 --> 00:11:03,000 So let's wait until this process, our video, then we can I will show you how you can work on this 184 00:11:03,000 --> 00:11:03,000 video. 185 00:11:03,000 --> 00:11:04,000 So. 186 00:11:06,000 --> 00:11:08,000 So big we can have one output. 187 00:11:08,000 --> 00:11:10,000 145 images from this video. 188 00:11:10,000 --> 00:11:19,000 If we get a one images one frame in each second like this video is around two minutes, 24 seconds. 189 00:11:19,000 --> 00:11:22,000 So it is around one 44 seconds. 190 00:11:22,000 --> 00:11:23,000 Okay. 191 00:11:23,000 --> 00:11:25,000 The total time span of this video is 144 second. 192 00:11:25,000 --> 00:11:29,000 So in each second, we are getting a single frame of image. 193 00:11:29,000 --> 00:11:29,000 Okay? 194 00:11:29,000 --> 00:11:33,000 So in this way, we are getting around 145 images. 195 00:11:33,000 --> 00:11:40,000 So okay, but as I have already enough sufficient dataset, so okay, so just choosing a high frame 196 00:11:40,000 --> 00:11:41,000 rate. 197 00:11:42,000 --> 00:11:46,000 Okay, but you can choose that one frame per second which is already appearing. 198 00:11:47,000 --> 00:11:48,000 Okay. 199 00:11:48,000 --> 00:11:49,000 So you can see this. 200 00:11:49,000 --> 00:11:57,000 We have some particle images which so we have converted this whole videos into small each frames of 201 00:11:57,000 --> 00:11:59,000 images you can see over here. 202 00:11:59,000 --> 00:12:03,000 So we can also click on Save and Continue. 203 00:12:03,000 --> 00:12:05,000 And annotate these images as well. 204 00:12:05,000 --> 00:12:09,000 And we can also add these images into our dataset as well. 205 00:12:09,000 --> 00:12:10,000 Okay. 206 00:12:10,000 --> 00:12:12,000 So I will be working on these images annotation. 207 00:12:12,000 --> 00:12:16,000 So I will assign these images for annotation to myself. 208 00:12:16,000 --> 00:12:19,000 So now we have two tasks. 209 00:12:19,000 --> 00:12:20,000 Currently. 210 00:12:21,000 --> 00:12:26,000 First, we need to annotate the 603 images which we have already uploaded from our system, downloaded 211 00:12:26,000 --> 00:12:30,000 from Kaggle, and these are the seven images which basically. 212 00:12:31,000 --> 00:12:32,000 Look at these images from YouTube. 213 00:12:32,000 --> 00:12:35,000 So I will annotate all these images. 214 00:12:35,000 --> 00:12:37,000 And let me start annotating this. 215 00:12:37,000 --> 00:12:43,000 And as the as the annotation completes, I will be get back over here and then we will see how we can 216 00:12:43,000 --> 00:12:49,000 export this data set and into the our CoLab notebook and run the training. 217 00:12:49,000 --> 00:12:51,000 So let me do the annotation and then I will back. 218 00:12:54,000 --> 00:12:55,000 Vice. 219 00:12:55,000 --> 00:12:55,000 I am back. 220 00:12:55,000 --> 00:12:57,000 I was able to annotate the images. 221 00:12:57,000 --> 00:13:01,000 So now our dataset consists of 665 images. 222 00:13:01,000 --> 00:13:07,000 Okay, so we have that divided our dataset into train valid and test split. 223 00:13:07,000 --> 00:13:13,000 The 70% of the images are in the training, while the 20% of the images are in the validation and 10% 224 00:13:13,000 --> 00:13:15,000 of the images are in the test set. 225 00:13:18,000 --> 00:13:21,000 Well, next, let's go to the generate section over here. 226 00:13:21,000 --> 00:13:28,000 And so we have total around 665 images and we have. 227 00:13:29,000 --> 00:13:30,000 Click on Continue. 228 00:13:30,000 --> 00:13:33,000 We have taught one class which is of pothole, okay? 229 00:13:33,000 --> 00:13:37,000 And the training set consists of 70% of the images which are for 65. 230 00:13:38,000 --> 00:13:43,000 Validation set consists of 133 images and testing set consists of 67 images. 231 00:13:43,000 --> 00:13:44,000 Okay. 232 00:13:44,000 --> 00:13:51,000 So you can also rebalance it like if you have very few images like 100 or 200 images in our data, in 233 00:13:51,000 --> 00:13:57,000 your data set, then you can make validation around zero and you can make such like 90% of your data 234 00:13:57,000 --> 00:14:04,000 for the training and 10% of the data or 10 to 15% of the data for the testing purpose and make validation. 235 00:14:04,000 --> 00:14:05,000 Zero. 236 00:14:05,000 --> 00:14:05,000 Okay. 237 00:14:05,000 --> 00:14:09,000 So you can also rebalance it if you want. 238 00:14:09,000 --> 00:14:09,000 Okay. 239 00:14:09,000 --> 00:14:15,000 So if here you can add any pre-processing step you want, like if you want to resize the image. 240 00:14:15,000 --> 00:14:22,000 I have resized my all the images into 640 plus 640 because the pre-trained model of YOLO v is trained 241 00:14:22,000 --> 00:14:25,000 on 640 cross 640 image dataset. 242 00:14:25,000 --> 00:14:27,000 So to get best of the results. 243 00:14:27,000 --> 00:14:32,000 So I have resized my images so to 640 cross 640 around. 244 00:14:32,000 --> 00:14:34,000 Okay then I will click on continue. 245 00:14:34,000 --> 00:14:37,000 So if my data set is low, less. 246 00:14:37,000 --> 00:14:44,000 If I want to increase my data set, then you can add augmentation step, which is around to flip your 247 00:14:45,000 --> 00:14:51,000 images in the data set, crop the images in the data set, rotate the images in the data set or saturate 248 00:14:52,000 --> 00:14:57,000 increase the brightness of the images in the data set so it will create duplicate images. 249 00:14:57,000 --> 00:14:59,000 For example, if I click on. 250 00:15:00,000 --> 00:15:02,000 But Abbott a rotation. 251 00:15:02,000 --> 00:15:02,000 Okay. 252 00:15:03,000 --> 00:15:09,000 So it will basically generate from a single image, it will generate two more images which are being 253 00:15:09,000 --> 00:15:12,000 rotated by -15 and plus 15 degree. 254 00:15:12,000 --> 00:15:14,000 Like you can set the rotation from here. 255 00:15:14,000 --> 00:15:15,000 Okay. 256 00:15:15,000 --> 00:15:18,000 So in this way, if you want, you can increase your data set size. 257 00:15:18,000 --> 00:15:23,000 If it is low by augmenting the data set, I will click on continue. 258 00:15:23,000 --> 00:15:23,000 Okay. 259 00:15:23,000 --> 00:15:26,000 So you can see that my data set size is increased. 260 00:15:26,000 --> 00:15:31,000 Like previously it was 665, now I have 1595. 261 00:15:31,000 --> 00:15:38,000 So using just a single augmentation of -15 degrees and plus 15 degree, which will generate two images 262 00:15:38,000 --> 00:15:39,000 from a single image. 263 00:15:39,000 --> 00:15:42,000 It has increased my data set by three times. 264 00:15:42,000 --> 00:15:42,000 Okay. 265 00:15:42,000 --> 00:15:45,000 It will also only increase the training data set. 266 00:15:45,000 --> 00:15:48,000 It will not increase your testing and validation data set. 267 00:15:48,000 --> 00:15:51,000 Please remember it will only increase your training data set. 268 00:15:51,000 --> 00:15:52,000 Then click on generate. 269 00:15:54,000 --> 00:15:59,000 It might take a few seconds, so please bear with me until this process gets complete. 270 00:16:01,000 --> 00:16:03,000 Okay, so I'm also waiting. 271 00:16:03,000 --> 00:16:05,000 So this process gets complete. 272 00:16:05,000 --> 00:16:12,000 So you can see that my training set images has increased because I have applied augmentation step and 273 00:16:12,000 --> 00:16:13,000 now it's 1300. 274 00:16:13,000 --> 00:16:16,000 While the validation and testing set, it hasn't changed. 275 00:16:16,000 --> 00:16:20,000 Okay, so I can start training from here as well. 276 00:16:20,000 --> 00:16:25,000 I can train also train on Roboflow, but it's better to train on the Google CoLab. 277 00:16:25,000 --> 00:16:27,000 So to export. 278 00:16:27,000 --> 00:16:30,000 So if you want to train on Roboflow, you can just click on Start Train. 279 00:16:30,000 --> 00:16:36,000 It will do the training and as the training completes, you will get an email that a training has completes 280 00:16:36,000 --> 00:16:39,000 and you can test your model over here as well on Roboflow as well. 281 00:16:39,000 --> 00:16:42,000 But we will test our model on Google CoLab. 282 00:16:42,000 --> 00:16:46,000 I'll always like some coding work, so I will click on export. 283 00:16:47,000 --> 00:16:50,000 I just like the format as YOLO v5 PyTorch format. 284 00:16:51,000 --> 00:16:52,000 They can continue. 285 00:16:53,000 --> 00:17:00,000 Basically, you can also select a PI torch format, but Ultralytics has also released YOLO V5 so you 286 00:17:00,000 --> 00:17:03,000 can select the YOLO V5 PI torch format as well. 287 00:17:03,000 --> 00:17:05,000 And just you need to copy this. 288 00:17:06,000 --> 00:17:13,000 From here and you will get be able to export this data set into your Google CoLab notebook. 289 00:17:13,000 --> 00:17:21,000 So in the next part, we will see how we can train our dataset, our custom model on path Potholes dataset 290 00:17:21,000 --> 00:17:22,000 in Google CoLab. 291 00:17:22,000 --> 00:17:25,000 So let's see you all in the next video tutorial.