1 00:00:03,000 --> 00:00:08,000 In this video tutorial, we will look at Multi-object tracking with Ultralytics YOLO. 2 00:00:09,000 --> 00:00:16,000 In object tracking, we assign a unique ID to each of the detected object, and then we track that detected 3 00:00:16,000 --> 00:00:19,000 object throughout the entire video frames. 4 00:00:19,000 --> 00:00:26,000 We will also plot the movement of detected object across multiple video frames by drawing the trails, 5 00:00:26,000 --> 00:00:31,000 which through which we can see what pattern the object is following. 6 00:00:31,000 --> 00:00:38,000 The applications of object tracking are limitless, from security and surveillance to real time sports 7 00:00:38,000 --> 00:00:39,000 analysis. 8 00:00:40,000 --> 00:00:47,000 In this tutorial for Object tracking for object detection, I will be using Ultralytics YOLO 11 model 9 00:00:47,000 --> 00:00:52,000 and for object tracking I will also use the Ultralytics package. 10 00:00:52,000 --> 00:00:59,000 Ultralytics package currently supports a two object tracking algorithms, one is Boardshort and addressed 11 00:00:59,000 --> 00:01:00,000 by Greg. 12 00:01:00,000 --> 00:01:07,000 So currently, uh, the politics package sports two object tracking algorithm, although there are other 13 00:01:07,000 --> 00:01:13,000 object tracking algorithms available as well, like Deepsort, sword and others as well. 14 00:01:13,000 --> 00:01:18,000 But currently the Ultralytics package, uh, supports two object tracking algorithm, which I told you 15 00:01:18,000 --> 00:01:20,000 one is called sort and other is by drag. 16 00:01:20,000 --> 00:01:26,000 So we will be using, uh, Ultralytics package to integrate object tracking with our object detection, 17 00:01:26,000 --> 00:01:27,000 uh, model. 18 00:01:27,000 --> 00:01:29,000 So let's get started with this. 19 00:01:29,000 --> 00:01:33,000 Uh, here you can see the demo is being played on your screen as well. 20 00:01:33,000 --> 00:01:39,000 In this demo, you can see that, uh, for each of the detected object, we have assigned a unique ID 21 00:01:39,000 --> 00:01:41,000 like one, three, nine, ten. 22 00:01:41,000 --> 00:01:43,000 And you can see that we are. 23 00:01:43,000 --> 00:01:50,000 So we are also, uh, checking the movement of the detected object by drawing the trace, which is telling 24 00:01:50,000 --> 00:01:54,000 us what path the object is being, uh, following. 25 00:01:54,000 --> 00:01:54,000 Okay. 26 00:01:54,000 --> 00:01:59,000 So this is a quick demo, and let's get started with this over here. 27 00:01:59,000 --> 00:02:07,000 So first of all, I will, uh, start writing down the code from scratch, and then I will show you 28 00:02:07,000 --> 00:02:11,000 how you can integrate object tracking with your object detection algorithm. 29 00:02:11,000 --> 00:02:17,000 So but before I go ahead, you can check this documentation by alternatives like how you can perform 30 00:02:17,000 --> 00:02:19,000 multi object tracking with Ultralytics YOLO. 31 00:02:19,000 --> 00:02:24,000 So we will be following this documentation uh in this complete tutorial. 32 00:02:24,000 --> 00:02:26,000 So you can also check out this documentation. 33 00:02:26,000 --> 00:02:31,000 And or you just need to search on Google by Multi Object Tracking with Ultralytics yolo. 34 00:02:31,000 --> 00:02:33,000 And you will just find the first link. 35 00:02:33,000 --> 00:02:37,000 And you can just click that link and you will be able to open this documentation. 36 00:02:37,000 --> 00:02:39,000 So let's start writing down the code. 37 00:02:40,000 --> 00:02:45,000 So over here you can see I have just created a project in PyCharm over here. 38 00:02:45,000 --> 00:02:48,000 And you can see I have multiple video files over here. 39 00:02:48,000 --> 00:02:53,000 And first of all I will just create a requirements.txt file. 40 00:03:01,000 --> 00:03:06,000 He had acquired the Ultralytics badge along with Ultralytics. 41 00:03:06,000 --> 00:03:12,000 Many other packages get installed as well, like OpenCV, Python, matplotlib, NumPy, Seaborn. 42 00:03:12,000 --> 00:03:14,000 So we will install. 43 00:03:19,000 --> 00:03:23,000 It install python r requirements dot txt. 44 00:03:25,000 --> 00:03:31,000 So now I'm just installing the latest package because in my requirements.txt file I have only mentioned 45 00:03:31,000 --> 00:03:32,000 the ultralytics package. 46 00:03:33,000 --> 00:03:35,000 R spellings are all for the requirements. 47 00:03:35,000 --> 00:03:36,000 That means. 48 00:03:53,000 --> 00:03:54,000 Let's run this. 49 00:03:54,000 --> 00:04:00,000 So if you have mentioned any other package in this requirements.txt file, this will also be get downloaded 50 00:04:00,000 --> 00:04:04,000 with the ultralytics package as well. 51 00:04:04,000 --> 00:04:09,000 So currently we only require ultralytics package along with ultralytics package like SciPy, torch, 52 00:04:09,000 --> 00:04:13,000 OpenCV, Python, NumPy and many other packages get installed as well. 53 00:04:13,000 --> 00:04:17,000 Like you can see matplotlib, OpenCV, Python over here. 54 00:04:17,000 --> 00:04:19,000 Okay so numpy. 55 00:04:20,000 --> 00:04:24,000 So you can see over here many other packages have been get installed pandas. 56 00:04:25,000 --> 00:04:30,000 So let's try to create a new.py file over here. 57 00:04:33,000 --> 00:04:35,000 So we'll write object. 58 00:04:38,000 --> 00:04:40,000 tracking.py. 59 00:04:44,000 --> 00:04:49,000 Or in the step number one we will import all the required. 60 00:04:51,000 --> 00:04:52,000 Libraries. 61 00:04:54,000 --> 00:04:56,000 So from Ultralytics. 62 00:04:58,000 --> 00:05:00,000 In port Yolo. 63 00:05:04,000 --> 00:05:08,000 We will load the YOLO in our model. 64 00:05:12,000 --> 00:05:18,000 So I will write model is equal to YOLO and we want the YOLO 11 nano model. 65 00:05:18,000 --> 00:05:20,000 Your 11 comes with five different models. 66 00:05:21,000 --> 00:05:27,000 YOLO 11 nano is the fastest, but it is the least accurate among other YOLO 11 models. 67 00:05:34,000 --> 00:05:37,000 The first one is tracking with it. 68 00:05:37,000 --> 00:05:39,000 All right. 69 00:05:40,000 --> 00:05:41,000 What sort? 70 00:05:43,000 --> 00:05:44,000 Of test. 71 00:05:44,000 --> 00:05:50,000 So if we just check out the documentation over here, uh, you will find that the default tracker is 72 00:05:50,000 --> 00:05:51,000 bot sort. 73 00:05:51,000 --> 00:05:51,000 Chart. 74 00:05:51,000 --> 00:05:55,000 Okay, so if you don't define that tracker over here like you right. 75 00:05:55,000 --> 00:06:00,000 Model or drag and don't define that tracker over here, let me just zoom in okay. 76 00:06:03,000 --> 00:06:03,000 Okay. 77 00:06:03,000 --> 00:06:09,000 And so if you right model or track but you don't define the tracker over here, it will use the default 78 00:06:09,000 --> 00:06:11,000 tracker which is what sort. 79 00:06:11,000 --> 00:06:16,000 So Ultralytics supports the following tracking algorithm about sort and by track. 80 00:06:16,000 --> 00:06:23,000 Uh so you can enable uh, any of these tracking algorithms like hidden sort or by track by, by passing 81 00:06:23,000 --> 00:06:27,000 the board sort dot yml file or by track dot yml file. 82 00:06:27,000 --> 00:06:33,000 So uh, by passing the arrangement yml configuration, you can activate any of the tracker. 83 00:06:33,000 --> 00:06:33,000 Okay. 84 00:06:35,000 --> 00:06:41,000 So if you want to check uh, the informations like uh, on the alternatives tracker base, like you 85 00:06:41,000 --> 00:06:44,000 can see it's post trackers, bot sort and by track. 86 00:06:44,000 --> 00:06:44,000 Okay. 87 00:06:46,000 --> 00:06:48,000 So let's move towards the code. 88 00:06:52,000 --> 00:06:54,000 Alright results model dot. 89 00:06:54,000 --> 00:06:58,000 So if I want to do object detection, I don't need to write anything. 90 00:06:58,000 --> 00:07:02,000 I just need to go over here and write the source in which I need to pass the video path. 91 00:07:02,000 --> 00:07:08,000 But if you want to add tracking, you just need to write track over here, and then you just need to 92 00:07:08,000 --> 00:07:13,000 define the source file like your input file over here. 93 00:07:16,000 --> 00:07:19,000 So I just need to go to the resources folder. 94 00:07:19,000 --> 00:07:21,000 Then I just need to go to the videos folder. 95 00:07:21,000 --> 00:07:27,000 And I can use video seven dot app. 96 00:07:31,000 --> 00:07:36,000 If I want to show this video I can write show is equal to true. 97 00:07:36,000 --> 00:07:41,000 And if you want to save the output video, uh, you can write save is equal to true. 98 00:07:41,000 --> 00:07:46,000 So shortly our packages will be get installed like it's doing indexing. 99 00:07:46,000 --> 00:07:48,000 So this will not take very much time. 100 00:07:48,000 --> 00:07:56,000 So let's wait for this to finish and then I will just write or implement this code. 101 00:07:56,000 --> 00:08:01,000 So now you can see currently we are using default tracker because I have not defined any tracker over 102 00:08:01,000 --> 00:08:02,000 here okay. 103 00:08:04,000 --> 00:08:07,000 So let's wait for the packages to finish. 104 00:08:12,000 --> 00:08:17,000 By now you can see that finally the packages package installation is being done. 105 00:08:17,000 --> 00:08:23,000 And now let's run this script over here okay, so the indexing is almost completed. 106 00:08:23,000 --> 00:08:25,000 Uh, let's wait for a few seconds. 107 00:08:26,000 --> 00:08:29,000 Uh, the indexing is being done, so this will not take very much time. 108 00:08:31,000 --> 00:08:33,000 Okay, let's run the object. 109 00:08:35,000 --> 00:08:40,000 So it takes around 5 to 6 minutes for the index package to get installed. 110 00:08:40,000 --> 00:08:44,000 Because the alternative is many other packages also get installed. 111 00:08:44,000 --> 00:08:46,000 So this takes some time okay. 112 00:08:46,000 --> 00:08:49,000 So But first of all, it will download the YOLO 11 nano model. 113 00:08:50,000 --> 00:08:51,000 It automatically downloads. 114 00:08:51,000 --> 00:08:58,000 You don't need to go to the GitHub repository and specifically download the YOLO 11 nano model. 115 00:08:58,000 --> 00:08:58,000 Okay? 116 00:09:15,000 --> 00:09:17,000 Okay, now you can see over here. 117 00:09:20,000 --> 00:09:24,000 Okay, so we are just getting issue no module named lab. 118 00:09:24,000 --> 00:09:24,000 Okay. 119 00:09:24,000 --> 00:09:26,000 So we can just write over here. 120 00:09:27,000 --> 00:09:30,000 Uh we also need to install the lab package as well. 121 00:09:30,000 --> 00:09:33,000 Um, actually it should include in their installation. 122 00:09:33,000 --> 00:09:36,000 Like you don't need to install it specifically. 123 00:09:40,000 --> 00:09:45,000 So now you can see I'm downloading installing the lab package over here. 124 00:09:45,000 --> 00:09:47,000 and you can see your 11 nano models being downloaded. 125 00:09:48,000 --> 00:09:49,000 So this. 126 00:09:49,000 --> 00:09:53,000 Installation will take two seconds more as well, so let's wait for it to get finished. 127 00:09:59,000 --> 00:10:01,000 Uh, we can ignore this. 128 00:10:01,000 --> 00:10:01,000 Uh. 129 00:10:06,000 --> 00:10:07,000 Uh, let me just check. 130 00:10:11,000 --> 00:10:13,000 Hopefully it should run now. 131 00:10:13,000 --> 00:10:15,000 And let's see how. 132 00:10:22,000 --> 00:10:24,000 It will perform an auto update over here. 133 00:10:34,000 --> 00:10:37,000 Okay, so here you can see we have the results. 134 00:10:37,000 --> 00:10:40,000 Now you can see that, uh, ID is equal to one. 135 00:10:40,000 --> 00:10:42,000 And here we are detecting the cars. 136 00:10:42,000 --> 00:10:43,000 So okay. 137 00:10:43,000 --> 00:10:49,000 but we are not able to assign the we are not able to do the reductions over here. 138 00:10:50,000 --> 00:10:51,000 Okay. 139 00:10:51,000 --> 00:10:55,000 So now you can see as the object comes here we are able to do the reductions. 140 00:10:55,000 --> 00:10:59,000 Now you can see that each of detected object is being assigned a unique ID. 141 00:10:59,000 --> 00:11:01,000 Now it looks pretty good. 142 00:11:01,000 --> 00:11:01,000 Yeah. 143 00:11:01,000 --> 00:11:06,000 So now you can see that each of the rejected object is being randomly assigned an ID. 144 00:11:06,000 --> 00:11:07,000 Okay. 145 00:11:07,000 --> 00:11:08,000 So if you use a. 146 00:11:08,000 --> 00:11:10,000 Currently I am using YOLO 11 nano model. 147 00:11:10,000 --> 00:11:17,000 But if you use uh, other models like YOLO 11 medium or your 11 x which have a high accuracy rate, 148 00:11:17,000 --> 00:11:19,000 so you will see that detection results will further improve. 149 00:11:19,000 --> 00:11:22,000 Currently I'm running it on my CPU machine. 150 00:11:22,000 --> 00:11:28,000 I don't have a GPU, so therefore I'm pressing YOLO 11 nano model because it is the fastest among YOLO 151 00:11:28,000 --> 00:11:30,000 11 model, but it is least accurate. 152 00:11:30,000 --> 00:11:37,000 Okay, so now you have over here you can see that we are able to detection and we are able to assign 153 00:11:37,000 --> 00:11:40,000 a unique ID to each of the detected object over here. 154 00:11:40,000 --> 00:11:41,000 Okay. 155 00:11:42,000 --> 00:11:49,000 So now you can see this is this object detection over here I am using 11 nano. 156 00:11:49,000 --> 00:11:52,000 So therefore there are some false positives over here as well. 157 00:11:52,000 --> 00:11:55,000 But this is what we do in object tracking. 158 00:11:55,000 --> 00:11:59,000 Like you can see we are assigning a unique ID to each of the detected object. 159 00:12:00,000 --> 00:12:00,000 Okay. 160 00:12:00,000 --> 00:12:07,000 And to stop this I will just click the pause button and it will stop okay. 161 00:12:07,000 --> 00:12:08,000 So. 162 00:12:11,000 --> 00:12:13,000 For next thing what we can do is. 163 00:12:15,000 --> 00:12:20,000 Now we will see how we can do tracking with by tracking okay. 164 00:12:26,000 --> 00:12:28,000 So we will just define. 165 00:12:30,000 --> 00:12:32,000 That is equal to model dot track. 166 00:12:32,000 --> 00:12:36,000 Over here force is equal to. 167 00:12:36,000 --> 00:12:38,000 Let's use some other video. 168 00:12:38,000 --> 00:12:41,000 I just need to go inside the resources folder. 169 00:12:41,000 --> 00:12:47,000 Then videos lets you create your eight dot board. 170 00:12:49,000 --> 00:12:56,000 I want to see the output and you can see over here, uh, our output is being saved over here for the 171 00:12:56,000 --> 00:12:58,000 previous, uh, run. 172 00:12:58,000 --> 00:12:59,000 Okay. 173 00:13:00,000 --> 00:13:04,000 So I want to save is equal to true. 174 00:13:08,000 --> 00:13:09,000 And we want to use. 175 00:13:12,000 --> 00:13:13,000 Byte pair dot. 176 00:13:13,000 --> 00:13:14,000 YML. 177 00:13:16,000 --> 00:13:22,000 And you like for the prediction in Yolo V8 model we have our different arguments. 178 00:13:22,000 --> 00:13:25,000 Similarly we have arguments for the tracking as well. 179 00:13:25,000 --> 00:13:29,000 Like we can set the constant ratio to 0.20. 180 00:13:29,000 --> 00:13:38,000 We can set the non-expression IOU to 0.5 so you can adjust these parameters. 181 00:13:39,000 --> 00:13:40,000 but we can make them too. 182 00:13:43,000 --> 00:13:45,000 So let's run this now. 183 00:13:46,000 --> 00:13:48,000 Okay, I made a mistake. 184 00:13:49,000 --> 00:13:51,000 I just need to comment this out. 185 00:14:08,000 --> 00:14:08,000 Over here. 186 00:14:08,000 --> 00:14:09,000 We have the output. 187 00:14:09,000 --> 00:14:14,000 Now, you can see over here, uh, we are able to detect the person over here in this video. 188 00:14:14,000 --> 00:14:22,000 And you can see that, uh, with each person, we have assigned a unique ID like seven, 6321. 189 00:14:22,000 --> 00:14:27,000 Like you can see, we are not able to detect these persons, but if you use, uh, other YOLO 11 models, 190 00:14:27,000 --> 00:14:29,000 currently I'm using 11 nano model. 191 00:14:29,000 --> 00:14:35,000 But if you use models, other YOLO 11 models which have a high accuracy rate, um, then it will be 192 00:14:35,000 --> 00:14:38,000 able to detect these persons over here as well. 193 00:14:38,000 --> 00:14:38,000 Okay. 194 00:14:38,000 --> 00:14:42,000 And but in this tutorial we are focusing on object tracking. 195 00:14:42,000 --> 00:14:43,000 Basically object tracking. 196 00:14:43,000 --> 00:14:46,000 We assign a unique ID to each of the detected object. 197 00:14:46,000 --> 00:14:47,000 Like you can see over here. 198 00:14:47,000 --> 00:14:50,000 We have assigned a unique ID and to each detected object. 199 00:14:50,000 --> 00:14:56,000 And then we can track that detected object throughout the entire video frames. 200 00:14:56,000 --> 00:14:57,000 Okay. 201 00:14:57,000 --> 00:14:59,000 So we are done with this. 202 00:15:00,000 --> 00:15:02,000 And that was pretty fine. 203 00:15:06,000 --> 00:15:06,000 Okay. 204 00:15:09,000 --> 00:15:13,000 So now first of all I will just comment all this out okay. 205 00:15:16,000 --> 00:15:19,000 So you can uncomment this and run this okay. 206 00:15:21,000 --> 00:15:30,000 So now I will just write a Python script using OpenCV, Python and YOLO 11 to run object tracking on 207 00:15:30,000 --> 00:15:31,000 video frames. 208 00:15:31,000 --> 00:15:35,000 Or you can also run on the live webcam feed as well. 209 00:15:35,000 --> 00:15:35,000 Okay. 210 00:15:38,000 --> 00:15:41,000 So let's add a separator. 211 00:15:43,000 --> 00:15:46,000 So we will write a Python script. 212 00:15:49,000 --> 00:15:53,000 Using OpenCV Python and. 213 00:15:59,000 --> 00:16:00,000 YOLO 11. 214 00:16:01,000 --> 00:16:10,000 Run object tracking on video frames. 215 00:16:10,000 --> 00:16:10,000 And. 216 00:16:15,000 --> 00:16:19,000 Okay so let's get started with this as well. 217 00:16:24,000 --> 00:16:29,000 So first of all will import all the required libraries. 218 00:16:31,000 --> 00:16:37,000 Import cv2 from I draw Netflix and thought YOLO. 219 00:16:40,000 --> 00:16:45,000 Then lower the YOLO 11 order right. 220 00:16:45,000 --> 00:16:52,000 Model Yolo Yolo 11 nano dot RT. 221 00:16:54,000 --> 00:17:04,000 Next we will create a video capture object like cv2 dot videocapture. 222 00:17:06,000 --> 00:17:10,000 We just need to go into the resources folder. 223 00:17:10,000 --> 00:17:16,000 Then we just need to go to the videos and we can use video seven. 224 00:17:16,000 --> 00:17:19,000 Or we can just use video five. 225 00:17:19,000 --> 00:17:21,000 Let's use some other video. 226 00:17:25,000 --> 00:17:36,000 Then we will go through very famous pipeline while True red form of frame cap dot read. 227 00:17:37,000 --> 00:17:44,000 So if we have able to read the video because that returns a boolean true, so it returns a that returns 228 00:17:44,000 --> 00:17:47,000 a boolean true that is able to read a video. 229 00:17:47,000 --> 00:17:48,000 Then. 230 00:17:52,000 --> 00:17:53,000 We will run. 231 00:17:56,000 --> 00:18:00,000 Event tracking on the video frames. 232 00:18:02,000 --> 00:18:03,000 We will write. 233 00:18:07,000 --> 00:18:11,000 Results is equal to model dot track. 234 00:18:14,000 --> 00:18:19,000 We have the input frame and we will set our argument assistant is equal to true. 235 00:18:23,000 --> 00:18:33,000 So this argument tells the tracker that it should expect a track from the previous frame or image in 236 00:18:33,000 --> 00:18:35,000 the current frame, so persistent is equal to true. 237 00:18:35,000 --> 00:18:40,000 Tells the tracker that it should expect tracks from the previous frame into the current frame. 238 00:18:40,000 --> 00:18:41,000 Okay. 239 00:18:46,000 --> 00:18:50,000 Well, next we will visualize the results on the frame. 240 00:18:55,000 --> 00:19:03,000 We have an updated frame is equal to results 0:00. 241 00:19:07,000 --> 00:19:10,000 Next we will display the annotated frame. 242 00:19:21,000 --> 00:19:24,000 We can write the name of the window as YOLO 11 tracking. 243 00:19:26,000 --> 00:19:32,000 So now in this window we will show short object tracking integrated with our object detection. 244 00:19:36,000 --> 00:19:37,000 All right. 245 00:19:37,000 --> 00:19:38,000 If it will not work. 246 00:19:38,000 --> 00:19:42,000 So we will add one millisecond delay between the consecutive frames. 247 00:19:43,000 --> 00:19:49,000 And if the users pass key, then the loop will break. 248 00:19:54,000 --> 00:19:56,000 We can add the command break the loop. 249 00:19:59,000 --> 00:20:02,000 If q key is pressed. 250 00:20:03,000 --> 00:20:04,000 Simple. 251 00:20:04,000 --> 00:20:10,000 And if it's unable to read the video frame, then it will break as well. 252 00:20:10,000 --> 00:20:16,000 And then it will release all the frames and we will destroy all the windows as well. 253 00:20:20,000 --> 00:20:23,000 So let's run this up now. 254 00:20:45,000 --> 00:20:49,000 So now over here, you can see that we are able to integrate object tracking. 255 00:20:49,000 --> 00:20:55,000 Like you can see that for each of the detected object, we are able to assign a unique ID like one with 256 00:20:55,000 --> 00:21:00,000 this person, six with this bicycle and eight with this bicycle. 257 00:21:00,000 --> 00:21:02,000 And the results look quite good. 258 00:21:02,000 --> 00:21:04,000 Like you can see with traffic lights. 259 00:21:04,000 --> 00:21:07,000 We are also able to assign a unique ID over here. 260 00:21:09,000 --> 00:21:12,000 Okay, so the results look quite promising to me. 261 00:21:13,000 --> 00:21:14,000 Okay. 262 00:21:16,000 --> 00:21:18,000 Yeah, the results look quite good. 263 00:21:18,000 --> 00:21:27,000 Like, we are able to assign a unique ID with each of the detected object, like 1715 Three 114. 264 00:21:27,000 --> 00:21:33,000 Okay, so if I press up, that loop will break that. 265 00:21:35,000 --> 00:21:44,000 So the next thing we can do is, um, now we'll see what, uh, path the object is following. 266 00:21:44,000 --> 00:21:48,000 So we'll draw the trail, or we can write another Python script over here. 267 00:21:54,000 --> 00:21:56,000 Object tracking trails. 268 00:22:03,000 --> 00:22:07,000 Obviously, plotting tracks over time. 269 00:22:12,000 --> 00:22:17,000 Or we will plot the movement of an object across multiple video frames. 270 00:22:21,000 --> 00:22:25,000 First of all, I will import all the required. 271 00:22:35,000 --> 00:22:36,000 Libraries. 272 00:22:53,000 --> 00:22:54,000 Are full. 273 00:22:54,000 --> 00:23:02,000 We will be saving the center point of each of the bounding box into our dictionary. 274 00:23:02,000 --> 00:23:06,000 And then using that center point, we will be drawing lines. 275 00:23:06,000 --> 00:23:06,000 Okay. 276 00:23:13,000 --> 00:23:19,000 So we will be saving the center point of the bounding box in each of the frame in our dictionary. 277 00:23:19,000 --> 00:23:21,000 And using that center point. 278 00:23:21,000 --> 00:23:22,000 Then we will draw the lines. 279 00:23:39,000 --> 00:23:41,000 So now we load the YOLO 11 model. 280 00:23:41,000 --> 00:23:45,000 Then we will create the video capture check. 281 00:23:53,000 --> 00:23:55,000 We just need to go inside the resources. 282 00:23:55,000 --> 00:23:57,000 Then we just need to go to the videos. 283 00:23:57,000 --> 00:24:00,000 And then we can use video five dot mp4. 284 00:24:02,000 --> 00:24:07,000 So we will store that track history into a dictionary. 285 00:24:24,000 --> 00:24:28,000 Then we will go through the video frames. 286 00:24:33,000 --> 00:24:34,000 Okay. 287 00:24:34,000 --> 00:24:35,000 App dot frame. 288 00:24:35,000 --> 00:24:38,000 So if we are able to read the video then. 289 00:24:47,000 --> 00:24:49,000 We'll run your 11. 290 00:24:53,000 --> 00:24:55,000 Tracking frame. 291 00:25:00,000 --> 00:25:02,000 Like model dot track. 292 00:25:03,000 --> 00:25:05,000 Here we have the source. 293 00:25:08,000 --> 00:25:09,000 Which is frame. 294 00:25:22,000 --> 00:25:29,000 Now we need to get the bounding box coordinates and the track IDs. 295 00:25:34,000 --> 00:25:35,000 Blind box. 296 00:25:35,000 --> 00:25:43,000 This is equal to because this will contain the bounding box coordinates for all objects in the frame. 297 00:25:43,000 --> 00:25:45,000 And then we'll loop through this so we can write. 298 00:25:46,000 --> 00:25:47,000 Result zero. 299 00:25:55,000 --> 00:25:56,000 New requirements. 300 00:25:56,000 --> 00:25:57,000 S center coordinate. 301 00:26:04,000 --> 00:26:06,000 Then we will also require the track ID. 302 00:26:06,000 --> 00:26:09,000 So we'll get the track IDs for all the objects in the frame. 303 00:26:33,000 --> 00:26:41,000 So similarly, in the previous.py file, we'll write the results on the frame as well. 304 00:26:44,000 --> 00:26:46,000 And I annotate the frame. 305 00:26:59,000 --> 00:27:03,000 So I can display the annotated frame as well. 306 00:27:19,000 --> 00:27:22,000 So then we can write if cv2 dot wait. 307 00:27:22,000 --> 00:27:23,000 Key one. 308 00:27:23,000 --> 00:27:32,000 There is a one millisecond delay between the consecutive frames and if the user press the W key. 309 00:27:39,000 --> 00:27:41,000 I missed one thing over here. 310 00:27:45,000 --> 00:27:47,000 Then the loop will break. 311 00:27:53,000 --> 00:27:59,000 So now I will release on the frame and destroy all the windows as well. 312 00:27:59,000 --> 00:28:02,000 So let me just run this. 313 00:28:18,000 --> 00:28:23,000 Now, currently you can see that we are able to implement object tracking, but we are not able to draw 314 00:28:23,000 --> 00:28:24,000 the tracks. 315 00:28:24,000 --> 00:28:26,000 We show the path the object is following. 316 00:28:27,000 --> 00:28:29,000 So I just press the W key. 317 00:28:29,000 --> 00:28:30,000 It will close. 318 00:28:30,000 --> 00:28:33,000 So we need to plot the tracks. 319 00:28:34,000 --> 00:28:37,000 For this I will write the code over here. 320 00:28:39,000 --> 00:28:42,000 So now I will loop through the center coordinates and the track IDs. 321 00:28:42,000 --> 00:28:52,000 So I will just write box format track ID and say here we have the boxes. 322 00:28:52,000 --> 00:28:53,000 So I drag. 323 00:28:55,000 --> 00:28:55,000 ID. 324 00:29:08,000 --> 00:29:12,000 Over here we have created a default dictionary but in the variable track history. 325 00:29:12,000 --> 00:29:14,000 Now it's time to use this. 326 00:29:20,000 --> 00:29:26,000 And this track ID, we have the object IDs for all the objects in the frame. 327 00:29:32,000 --> 00:29:38,000 I submit that each track ID will save the center point of the bounding box, and we will then connect 328 00:29:38,000 --> 00:29:41,000 the center point of the bounding box for all the frames. 329 00:29:41,000 --> 00:29:46,000 And then we will draw the lines which tell us the path the object is following. 330 00:30:17,000 --> 00:30:23,000 So we need to obtain that points or center points for the 30 frames. 331 00:30:23,000 --> 00:30:23,000 Okay. 332 00:30:24,000 --> 00:30:26,000 So we write the ten track. 333 00:30:29,000 --> 00:30:29,000 30. 334 00:30:30,000 --> 00:30:34,000 So we will retain the center points for the 30 frames. 335 00:30:34,000 --> 00:30:41,000 So if you increase this value to 60 or 90 uh the the tracking trail line will increase. 336 00:30:41,000 --> 00:30:41,000 Okay. 337 00:30:41,000 --> 00:30:43,000 I will just show you as well. 338 00:30:52,000 --> 00:30:54,000 Now don't draw that tracking lines. 339 00:31:18,000 --> 00:31:23,000 Or using Siri to dot polling lines will draw the trails or the lines. 340 00:31:30,000 --> 00:31:32,000 Here we have the information or the center points. 341 00:31:43,000 --> 00:31:46,000 And the thickness will be ten. 342 00:31:46,000 --> 00:31:50,000 Okay, so that works pretty fine. 343 00:31:50,000 --> 00:31:51,000 Let's run this up. 344 00:31:51,000 --> 00:31:55,000 Now, uh, we can just use video. 345 00:31:56,000 --> 00:31:59,000 Uh, okay. 346 00:32:02,000 --> 00:32:06,000 So let's see how does it works. 347 00:32:09,000 --> 00:32:14,000 The color of these lines will be a white because we have just set the color to white. 348 00:32:14,000 --> 00:32:17,000 You can just change the color to any other color as well. 349 00:32:18,000 --> 00:32:23,000 Okay, so now you can see in the white we have the trails over here. 350 00:32:23,000 --> 00:32:26,000 So we chose the path the object is following. 351 00:32:26,000 --> 00:32:27,000 Okay. 352 00:32:28,000 --> 00:32:32,000 So now you can see over here we have the trails over here. 353 00:32:33,000 --> 00:32:38,000 Uh so let me just show you with other video. 354 00:32:39,000 --> 00:32:39,000 Video. 355 00:32:39,000 --> 00:32:45,000 So and we can just change the color to yellow. 356 00:32:49,000 --> 00:32:50,000 Let's run this now. 357 00:33:05,000 --> 00:33:06,000 Okay. 358 00:33:06,000 --> 00:33:08,000 So here we got the error call. 359 00:33:09,000 --> 00:33:09,000 Yeah. 360 00:33:16,000 --> 00:33:18,000 So how we can deal with this? 361 00:33:20,000 --> 00:33:21,000 We can write. 362 00:33:24,000 --> 00:33:26,000 XYZ0. 363 00:33:28,000 --> 00:33:30,000 But not id. 364 00:33:34,000 --> 00:33:38,000 Is not null. 365 00:33:51,000 --> 00:33:52,000 Okay. 366 00:33:53,000 --> 00:33:56,000 So now hopefully this will work out. 367 00:34:02,000 --> 00:34:03,000 So let's see. 368 00:34:12,000 --> 00:34:17,000 Now over here you can see that we are drawing the trails with a blue color over here. 369 00:34:17,000 --> 00:34:20,000 Okay, okay. 370 00:34:28,000 --> 00:34:34,000 Okay, so now you can see that these trails are telling us the pattern the object is following. 371 00:34:39,000 --> 00:34:39,000 Okay. 372 00:34:39,000 --> 00:34:45,000 So now you can see over here we are drawing the trails over here as well. 373 00:34:52,000 --> 00:34:54,000 So if I just press the W key this will close. 374 00:34:54,000 --> 00:34:59,000 And if I just write 90 over here let me show you what that does. 375 00:34:59,000 --> 00:34:59,000 It makes. 376 00:35:19,000 --> 00:35:24,000 So if I just write 90 over here, the length of these trails when further increase. 377 00:35:25,000 --> 00:35:25,000 Okay. 378 00:35:25,000 --> 00:35:32,000 So now you can see that like basically this is telling the path uh, the object is being following. 379 00:35:32,000 --> 00:35:38,000 Now you can see that the length has how much increased okay. 380 00:35:41,000 --> 00:35:43,000 So this is uh, quite makes impact. 381 00:35:44,000 --> 00:35:46,000 But 30 is the ideal value. 382 00:35:46,000 --> 00:35:46,000 Okay. 383 00:35:46,000 --> 00:35:52,000 So now we have seen that how we can plot the movement of object, how we can integrate object tracking. 384 00:35:52,000 --> 00:35:54,000 Now we will implement multi threading. 385 00:35:54,000 --> 00:35:56,000 Multi threaded tracking. 386 00:35:56,000 --> 00:36:02,000 So multi threaded tracking provides us the capability to run object tracking on multiple video streams 387 00:36:02,000 --> 00:36:06,000 or multiple live cam webcam feeds as well. 388 00:36:06,000 --> 00:36:09,000 Okay, so we can create another.py file. 389 00:36:23,000 --> 00:36:26,000 For multi threaded. 390 00:36:29,000 --> 00:36:31,000 Tracking provides. 391 00:36:34,000 --> 00:36:34,000 The. 392 00:36:37,000 --> 00:36:47,000 Capability to run check tracking on multiple video frames. 393 00:36:47,000 --> 00:36:56,000 Okay so first we will import all the required libraries. 394 00:37:03,000 --> 00:37:04,000 Import trading. 395 00:37:06,000 --> 00:37:11,000 Import cv2 from Ultra LaTeX. 396 00:37:12,000 --> 00:37:13,000 Import your. 397 00:37:16,000 --> 00:37:23,000 So next we need to define model names like what different models we use need to use in the multiple 398 00:37:23,000 --> 00:37:29,000 video streams and video sources. 399 00:37:29,000 --> 00:37:32,000 You can also define the webcam sources as well. 400 00:37:39,000 --> 00:37:48,000 We had windows in M1 nano model, and we will also use your 11 nano segmentation model as well. 401 00:37:53,000 --> 00:38:02,000 So we have sources equal to just need to go inside the resources, then there will need to go to the 402 00:38:02,000 --> 00:38:06,000 videos and first we'll run video seven dot mp4. 403 00:38:06,000 --> 00:38:09,000 Then we have sources. 404 00:38:12,000 --> 00:38:18,000 Videos video eight dot mp4 okay. 405 00:38:23,000 --> 00:38:27,000 So now we will run YOLO 11 tracker in its own track for concurrent processing. 406 00:38:27,000 --> 00:38:29,000 So we'll create a function. 407 00:38:32,000 --> 00:38:32,000 Okay. 408 00:38:34,000 --> 00:38:43,000 Run tracker that we just need to pass the model name and the file name in the input. 409 00:38:47,000 --> 00:38:49,000 So we'll run YOLO 11 tracker in its. 410 00:38:52,000 --> 00:38:53,000 Own thread. 411 00:38:55,000 --> 00:38:57,000 For concurrent processing. 412 00:39:10,000 --> 00:39:12,000 So we'll pass the model name to the. 413 00:39:12,000 --> 00:39:13,000 Over here. 414 00:39:15,000 --> 00:39:21,000 Then we'll write results model dot back because we need to implement bracket. 415 00:39:22,000 --> 00:39:25,000 Then in the source over here we have the file name. 416 00:39:30,000 --> 00:39:32,000 We need to save the output video. 417 00:39:36,000 --> 00:39:40,000 Then we need in the documentation they have set stream is equal to true. 418 00:39:40,000 --> 00:39:43,000 But this doesn't show the output video. 419 00:39:43,000 --> 00:39:44,000 So we just need to write. 420 00:39:44,000 --> 00:39:46,000 Show is equal to true. 421 00:39:48,000 --> 00:39:50,000 Then we will loop through the results. 422 00:39:57,000 --> 00:40:01,000 Now create and start tracker thread. 423 00:40:07,000 --> 00:40:11,000 I will now create and start tracker threads using a for loop. 424 00:40:35,000 --> 00:40:38,000 For now I will loop through the video files and the model names. 425 00:40:58,000 --> 00:41:00,000 So the arguments will be. 426 00:41:04,000 --> 00:41:06,000 Modern name and the video file. 427 00:41:27,000 --> 00:41:32,000 So now we will run YOLO tracker in its own thread for concurrent processing. 428 00:41:32,000 --> 00:41:34,000 I will show you in the output how this works. 429 00:41:39,000 --> 00:41:41,000 Let's start this up now. 430 00:41:45,000 --> 00:41:52,000 Now before it completes for example, if any video completes, it will wait for all the tracker threads 431 00:41:52,000 --> 00:41:53,000 to finish. 432 00:42:15,000 --> 00:42:19,000 I will clean up and close all video files. 433 00:42:22,000 --> 00:42:24,000 So now it's clean up and close all windows. 434 00:42:24,000 --> 00:42:27,000 So we write cv2 dot destroy all windows. 435 00:42:27,000 --> 00:42:29,000 So this is how it works. 436 00:42:30,000 --> 00:42:32,000 And let's run this and see if it works. 437 00:42:44,000 --> 00:42:47,000 So we have 11 nano model but we don't have the segmentation model. 438 00:42:47,000 --> 00:42:48,000 So it's downloading. 439 00:42:49,000 --> 00:42:50,000 And let's. 440 00:42:52,000 --> 00:43:00,000 So this is our first video and here is our second video. 441 00:43:00,000 --> 00:43:00,000 Okay. 442 00:43:02,000 --> 00:43:09,000 So you can see over here we are able to implement multithreading. 443 00:43:11,000 --> 00:43:16,000 So using multithreading it provides us the capability to run object tracking on multiple video streams 444 00:43:16,000 --> 00:43:18,000 like you can see over here. 445 00:43:18,000 --> 00:43:22,000 On one side I am implementing object tracking with the segmentation model. 446 00:43:22,000 --> 00:43:29,000 And on the other side on the left hand side I am implementing object tracking with the object detection 447 00:43:29,000 --> 00:43:30,000 model. 448 00:43:30,000 --> 00:43:30,000 Okay. 449 00:43:30,000 --> 00:43:33,000 So now you can see that the results look pretty good. 450 00:43:33,000 --> 00:43:36,000 Like you can clearly see over here. 451 00:43:36,000 --> 00:43:40,000 Uh, the results look quite promising to me as well. 452 00:43:40,000 --> 00:43:41,000 Okay. 453 00:43:41,000 --> 00:43:47,000 So now uh, using a multithreading we are able to implement object tracking. 454 00:43:48,000 --> 00:43:51,000 So that is results look quite promising. 455 00:43:51,000 --> 00:43:56,000 So now we are able to implement object tracking on multiple video streams as well. 456 00:44:03,000 --> 00:44:09,000 So multi threaded tracking provides us the capability to run object tracking on multiple video streams 457 00:44:09,000 --> 00:44:11,000 which you can clearly see over here. 458 00:44:11,000 --> 00:44:14,000 And over here you can see that we are doing concurrent processing. 459 00:44:14,000 --> 00:44:20,000 We are running video seven and video eight are in parallel over here. 460 00:44:20,000 --> 00:44:25,000 You can also run on live webcam feeds like if you have multiple feeds coming, you can run it on multiple 461 00:44:25,000 --> 00:44:26,000 feeds as well. 462 00:44:26,000 --> 00:44:29,000 And you can see this is what we call concurrent processing. 463 00:44:29,000 --> 00:44:35,000 Like we are running video seven and video eight concurrently at the same time. 464 00:44:35,000 --> 00:44:38,000 Okay, okay. 465 00:44:38,000 --> 00:44:39,000 So this is how it works. 466 00:44:39,000 --> 00:44:42,000 And using multiple threading tracking. 467 00:44:43,000 --> 00:44:46,000 Uh, we have run object tracking on multiple video streams. 468 00:44:46,000 --> 00:44:47,000 That's all from this tutorial. 469 00:44:47,000 --> 00:44:49,000 Thank you for watching.