1 00:00:02,450 --> 00:00:10,580 In this tutorial, we will learn how we can do object tracking using YOLO v9 and sort algorithm for 2 00:00:10,580 --> 00:00:11,660 object detection. 3 00:00:11,660 --> 00:00:18,830 I will be using the Yolo v9 algorithm and to track each of the detected object I will be using the sort 4 00:00:18,830 --> 00:00:19,910 algorithm. 5 00:00:19,910 --> 00:00:21,740 So let's get started with it. 6 00:00:21,740 --> 00:00:27,860 Over here you can see that I have created a new project by the name Yolo v nine sort recording, and 7 00:00:27,860 --> 00:00:33,620 I have just created this project in my local directory, uh, in my local directory so you can create 8 00:00:33,620 --> 00:00:39,470 a project, uh, new project in any of your local directory as well, using PyCharm or any other IDE 9 00:00:39,500 --> 00:00:40,250 as well. 10 00:00:40,250 --> 00:00:46,250 So I have listed all the steps which I will be following in this tutorial in this notepad. 11 00:00:46,250 --> 00:00:51,110 So over here in the step number one we will clone the Yolo v nine GitHub repo. 12 00:00:51,110 --> 00:00:54,800 So I will just go over here in the browser. 13 00:00:54,800 --> 00:00:57,440 So here you can see the Yolo v nine GitHub repo. 14 00:00:57,440 --> 00:01:02,870 So to clone it I will just click over here and copy the URL to the clipboard. 15 00:01:02,870 --> 00:01:03,230 Okay. 16 00:01:03,230 --> 00:01:04,340 So now it's copied. 17 00:01:04,340 --> 00:01:07,580 And now I will just go back to the PyCharm from here. 18 00:01:07,580 --> 00:01:11,930 And I will just open the terminal terminal over here. 19 00:01:11,930 --> 00:01:18,530 And to clone this complete repo GitHub repo into my this into this project yolo v nine sort recording 20 00:01:18,530 --> 00:01:20,330 I will just write it clone. 21 00:01:21,600 --> 00:01:25,500 And I will add the URL of the benign repo over here. 22 00:01:25,500 --> 00:01:29,550 So now you can see that the cloning has started and it's completed. 23 00:01:29,550 --> 00:01:33,510 So now I will if I just refresh this over here. 24 00:01:34,760 --> 00:01:35,900 Okay okay. 25 00:01:35,900 --> 00:01:37,160 So no no need to refresh. 26 00:01:37,160 --> 00:01:39,380 We have this yolo v nine wrapper over here. 27 00:01:39,380 --> 00:01:45,950 So now I will just set this yolo v9 wrapper which have cloned into my project as my current directory. 28 00:01:45,950 --> 00:01:49,130 So I will just set this YOLO v nine folder as my current directory. 29 00:01:49,130 --> 00:01:52,310 And you can see we have all the files over here as well. 30 00:01:52,310 --> 00:01:57,650 So to set this as YOLO benign folder as my current directory I will write CD which stands for current 31 00:01:57,650 --> 00:02:00,500 directory and I will just write YOLO v nine over here. 32 00:02:00,500 --> 00:02:04,430 So this YOLO v nine folder is being set as my current directory. 33 00:02:04,430 --> 00:02:09,950 So in the step number one uh we need to install all the required packages. 34 00:02:09,950 --> 00:02:16,730 So over here in the requirements.txt file you can find all the required packages that are required to 35 00:02:16,730 --> 00:02:17,480 run this. 36 00:02:18,820 --> 00:02:21,670 Uh, to run that detection and tracking script successfully. 37 00:02:21,670 --> 00:02:28,510 So which include OpenCV, Python, matplotlib, NumPy, SciPy, Torchvision, Tqdm and pandas, TensorFlow, 38 00:02:28,510 --> 00:02:33,880 js and all the required packages are listed over here, so these packages are required to run the detection 39 00:02:33,880 --> 00:02:35,800 and tracking script successfully. 40 00:02:35,800 --> 00:02:40,330 Or if you want to train your object detection model on any custom data set, you need to install all 41 00:02:40,330 --> 00:02:43,570 these packages so we can install all these packages one by one. 42 00:02:43,570 --> 00:02:46,990 Or there is another way we can do which is quite easy way. 43 00:02:46,990 --> 00:02:51,760 So you can just write over here pip install minus r. 44 00:02:52,980 --> 00:02:54,090 Required. 45 00:02:55,230 --> 00:02:56,640 Men's dot txt. 46 00:02:56,670 --> 00:03:03,360 So when you write pip install r requirements.txt it is basically install all the packages that are required 47 00:03:03,360 --> 00:03:05,250 listed in the requirements.txt file. 48 00:03:05,340 --> 00:03:10,020 So it will install all these packages that are listed in this requirements.txt file. 49 00:03:10,020 --> 00:03:12,240 So I will just click on enter from here. 50 00:03:12,240 --> 00:03:17,340 And now all the packages that are listed inside the requirements.txt file will get installed. 51 00:03:29,700 --> 00:03:34,500 So this might take a few more seconds for the process to get completed and all these packages to get 52 00:03:34,500 --> 00:03:34,980 installed. 53 00:03:34,980 --> 00:03:39,030 You can see that the packages installation is currently in progress. 54 00:03:43,120 --> 00:03:47,740 Now you can see over here all the required packages are being installed successfully. 55 00:03:47,770 --> 00:03:49,870 Next if I just open this up. 56 00:03:49,870 --> 00:03:55,270 So now we need to download a sample video from the Google Drive link below, or download a sample image 57 00:03:55,270 --> 00:03:56,440 from the Google Drive link below. 58 00:03:56,440 --> 00:03:58,780 So I will just copy this link from here. 59 00:03:59,050 --> 00:04:06,430 So to download any image or a video from the Google Drive link, we need to install the G down uh library 60 00:04:06,430 --> 00:04:06,910 as well. 61 00:04:06,910 --> 00:04:09,340 So first I will install the G down library. 62 00:04:09,730 --> 00:04:12,070 So I will just write pip install g down. 63 00:04:12,070 --> 00:04:13,780 So this will install this package. 64 00:04:13,780 --> 00:04:20,770 So this help us to download uh any video image or any file from the or any zip file or any other file 65 00:04:20,770 --> 00:04:26,410 from the Google drive directly into this, our local IDE, or either in Google Colab, notebook, Jupyter 66 00:04:26,410 --> 00:04:29,920 Notebook, Visual Studio or PyCharm or any other IDE as well. 67 00:04:29,920 --> 00:04:32,830 So g down stand for Google Drive download. 68 00:04:32,860 --> 00:04:39,640 I like this thing, so I will just uh, plus I will add this up in the requirements.txt file as well. 69 00:04:39,640 --> 00:04:43,780 So in any case, when you install the requirements dot txt file. 70 00:04:43,780 --> 00:04:47,320 So this package also gets stored along with this as well. 71 00:04:47,320 --> 00:04:49,240 So I g down here as well. 72 00:04:51,110 --> 00:04:51,590 Okay. 73 00:04:51,590 --> 00:04:58,160 So next I will just open this up, copy this link, and I will just add this link over here and click 74 00:04:58,160 --> 00:04:58,730 on enter. 75 00:04:58,730 --> 00:05:04,160 So this will download a sample image from the drive into this our project. 76 00:05:04,160 --> 00:05:06,200 So over here image one dot jpg. 77 00:05:06,200 --> 00:05:07,910 So if I let me see. 78 00:05:07,910 --> 00:05:11,630 So here you can see we have the image one dot jpg get installed. 79 00:05:11,630 --> 00:05:12,710 So. 80 00:05:16,180 --> 00:05:19,270 Next, we need to download sample videos from the Google Drive link below. 81 00:05:19,270 --> 00:05:23,170 So first I will just copy this Google Drive link and enter. 82 00:05:23,170 --> 00:05:27,640 So this will download a video from the drive into this PyCharm IDE. 83 00:05:31,350 --> 00:05:35,220 So test dot mp4 is the name of the demo video. 84 00:05:35,250 --> 00:05:36,840 So over here you can see. 85 00:05:37,900 --> 00:05:43,030 We have the test walked MP4 video over here next if I just go over here. 86 00:05:43,030 --> 00:05:46,930 So when you install the second demo video over here. 87 00:05:46,930 --> 00:05:49,150 So I will just add this link over here. 88 00:05:52,240 --> 00:05:55,480 So this will take a few more seconds before we have it. 89 00:05:55,510 --> 00:05:58,090 So now you can see we have the test three dot mp4. 90 00:05:58,120 --> 00:05:59,650 This is the other demo video. 91 00:05:59,650 --> 00:06:03,130 And if I just refresh this up here. 92 00:06:05,610 --> 00:06:09,450 So you can see here we have test three dot mp4 test dot mp4. 93 00:06:09,450 --> 00:06:15,090 So we have uh downloaded the sample image and sample videos from the Google Drive directly into this 94 00:06:15,090 --> 00:06:17,010 PyCharm IDE as well. 95 00:06:17,010 --> 00:06:19,320 So that's been done successfully. 96 00:06:19,320 --> 00:06:24,390 So download the sample image, download the sample videos from Google Drive link below. 97 00:06:24,390 --> 00:06:26,850 So now we need to do object detection on image. 98 00:06:26,850 --> 00:06:30,870 So as we uh first uh to do object detection on image. 99 00:06:30,870 --> 00:06:35,490 So you just need to go into this yolo v nine GitHub repo over here. 100 00:06:35,640 --> 00:06:38,940 Uh you will check the releases version 0.1. 101 00:06:38,940 --> 00:06:42,930 And over here you will see the model weights file over here. 102 00:06:42,930 --> 00:06:44,850 So you can just click over here. 103 00:06:44,850 --> 00:06:47,010 And the model weights file will be downloaded. 104 00:06:47,010 --> 00:06:49,470 So if you just click over here. 105 00:06:50,890 --> 00:06:52,630 So this will download the model weights. 106 00:06:52,750 --> 00:06:58,240 So I will just cancel this out because I have already have downloaded the model weights. 107 00:06:58,240 --> 00:06:59,740 And let me show you. 108 00:06:59,740 --> 00:07:04,840 So over here you can see over here we are the model weights. 109 00:07:05,020 --> 00:07:07,630 So here you can see we have the model weights. 110 00:07:07,630 --> 00:07:08,680 So. 111 00:07:11,720 --> 00:07:18,890 Copy this file name over here, and I will just go into my project over here, and I will just add the 112 00:07:18,890 --> 00:07:20,720 weights file over here as well. 113 00:07:20,720 --> 00:07:23,750 So you can see we have the model weights over here. 114 00:07:24,200 --> 00:07:25,760 So that's been done. 115 00:07:26,150 --> 00:07:29,570 So next we will just go to the code from here. 116 00:07:30,800 --> 00:07:32,150 And over here. 117 00:07:32,150 --> 00:07:35,330 So if you just go below down below. 118 00:07:35,330 --> 00:07:40,670 So to do inference with YOLO v nine, like if you want to do object detection on image with yolo v nine, 119 00:07:40,940 --> 00:07:46,100 uh, by using the pre-trained model weights, you will we will using detect test.py script file and 120 00:07:46,100 --> 00:07:47,960 in the source will pass the image path. 121 00:07:47,960 --> 00:07:51,740 And then we have the pre-trained model weights over here. 122 00:07:51,740 --> 00:07:52,730 So that's all. 123 00:07:52,730 --> 00:07:55,550 So I will just go over here. 124 00:07:57,390 --> 00:07:59,310 I will just write Python. 125 00:08:00,340 --> 00:08:02,590 You take that dashed.py. 126 00:08:02,590 --> 00:08:06,430 So here we have the detect-dual.py file. 127 00:08:06,580 --> 00:08:14,770 And then we have the source as image one dot jpg okay. 128 00:08:14,770 --> 00:08:17,590 And then we have the weights as. 129 00:08:18,410 --> 00:08:23,270 Yolo b9 that c dot b2. 130 00:08:23,480 --> 00:08:26,810 So you can see here we have the your weights file over here. 131 00:08:26,810 --> 00:08:29,480 Here we have the input image file over here. 132 00:08:29,480 --> 00:08:31,580 And if I just show you the input image. 133 00:08:31,580 --> 00:08:32,750 So this is our input image. 134 00:08:32,750 --> 00:08:35,780 And we will be doing object detection on this input image. 135 00:08:36,550 --> 00:08:37,030 Okay. 136 00:08:37,030 --> 00:08:41,350 Plus I don't have the GPU in my local system, so I will be using CPU. 137 00:08:41,350 --> 00:08:45,340 So device, CPU and um, anything else. 138 00:08:45,340 --> 00:08:48,100 So I want to see the output image. 139 00:08:48,100 --> 00:08:50,140 So I will just write dash dash view image. 140 00:08:50,140 --> 00:08:53,200 Plus I will just put it-dual.py. 141 00:08:53,200 --> 00:08:56,200 And I will just make one change in the code. 142 00:08:59,040 --> 00:09:01,980 I will just change it to zero. 143 00:09:01,980 --> 00:09:07,110 So the output will be image will be displayed for infinite time if I just write one. 144 00:09:07,110 --> 00:09:12,000 So the output image will be displayed for one millisecond, which is very small and we will not be able 145 00:09:12,000 --> 00:09:13,020 to see the results. 146 00:09:13,020 --> 00:09:16,020 So now it will be displayed for infinite time. 147 00:09:16,020 --> 00:09:18,960 Okay, so let's run this up now. 148 00:09:20,070 --> 00:09:21,780 I hope it will work fine. 149 00:09:21,780 --> 00:09:26,250 But if in any case, in theory, if there is any error, we will try to fix this up. 150 00:09:26,990 --> 00:09:31,400 So now we are doing object detection on this input image. 151 00:09:31,580 --> 00:09:32,930 Image one dot jpg. 152 00:09:33,080 --> 00:09:37,670 And let's see whether we are able to detect different objects inside this image or not. 153 00:09:39,430 --> 00:09:42,520 I'm using CPU so this will take some time. 154 00:09:42,520 --> 00:09:48,280 Like with GPU, the inference will be very fast, but with GPU the but with CPU the inference is very 155 00:09:48,280 --> 00:09:48,700 slow. 156 00:09:48,700 --> 00:09:53,260 So over here you can see that we are able to detect burst persons or traffic lights. 157 00:09:53,260 --> 00:09:56,320 Like you can see the persons are bit far away and very blurred. 158 00:09:56,320 --> 00:09:58,960 But the Yolov5 model is able to detect successfully. 159 00:09:58,960 --> 00:10:02,020 We are able to like bicycle backpack, hand bags. 160 00:10:02,170 --> 00:10:04,930 So the results look quite impressive from here. 161 00:10:05,140 --> 00:10:05,710 Okay. 162 00:10:05,710 --> 00:10:07,390 And bag is also detected. 163 00:10:07,390 --> 00:10:08,620 So. 164 00:10:09,490 --> 00:10:13,720 Okay, so if you can press any key, the, uh, it will the script will close up and the output image 165 00:10:13,720 --> 00:10:14,920 will also be closed. 166 00:10:15,220 --> 00:10:21,070 So next we will see how we can do, uh, object detection on a video. 167 00:10:21,070 --> 00:10:21,580 So. 168 00:10:22,570 --> 00:10:24,730 Let's make it to one again. 169 00:10:24,730 --> 00:10:27,490 So each frame will be displayed for one millisecond. 170 00:10:27,490 --> 00:10:32,680 And over here I will just add the path for the video. 171 00:10:32,680 --> 00:10:34,870 So over here I will just write. 172 00:10:34,960 --> 00:10:35,350 Um. 173 00:10:36,710 --> 00:10:40,160 Best free DAW MP. 174 00:10:41,880 --> 00:10:47,040 So now we'll be doing object detection on this street MP4 video. 175 00:10:47,040 --> 00:10:49,350 And let's see how does it goes. 176 00:10:50,780 --> 00:10:53,120 So we have done object detection on video. 177 00:10:53,120 --> 00:10:55,790 We have done now we are doing object detection on. 178 00:10:55,790 --> 00:10:57,530 We have done object detection on image. 179 00:10:57,530 --> 00:10:59,810 Now we are doing object detection on video. 180 00:10:59,810 --> 00:11:04,580 And next we will be doing object detection object tracking using yolo v9 and sort algorithm. 181 00:11:04,670 --> 00:11:08,330 So currently we are doing object detection on video. 182 00:11:09,920 --> 00:11:10,070 Okay. 183 00:11:10,070 --> 00:11:10,730 So. 184 00:11:11,090 --> 00:11:11,330 So. 185 00:11:13,180 --> 00:11:13,630 Okay. 186 00:11:14,200 --> 00:11:17,800 So over here you can see that we are able to detect the cars. 187 00:11:18,160 --> 00:11:19,930 Uh, so these are the false positives. 188 00:11:19,930 --> 00:11:21,250 This is the false positive. 189 00:11:21,250 --> 00:11:24,280 And here you can see that we are able to detect the cars. 190 00:11:24,310 --> 00:11:29,590 So here you can see we are also able to detect the cars as well. 191 00:11:29,650 --> 00:11:34,360 So the detection is very slow because we are running it on CPU machine. 192 00:11:34,360 --> 00:11:38,140 But if you use GPU the detections uh the frame rate will be very high. 193 00:11:38,170 --> 00:11:40,090 Currently the frame rates are very low. 194 00:11:40,180 --> 00:11:45,880 Uh, but if you test it on GPU, the frame rate will be around 5 to 6 fps frame per second. 195 00:11:45,880 --> 00:11:49,720 But I'm testing on CPU so the frame rate is very low. 196 00:11:51,590 --> 00:11:55,310 But you can see over here we are able to track the cars very successfully. 197 00:11:55,310 --> 00:11:56,630 And these three. 198 00:11:56,750 --> 00:11:59,150 And this truck is pos positive. 199 00:12:04,030 --> 00:12:08,110 So false positive means like there is no car, but it's detecting the car. 200 00:12:08,110 --> 00:12:10,060 So this is a false positive prediction. 201 00:12:10,150 --> 00:12:15,520 Like there doesn't exist any car over here and there doesn't exist any truck, but it's returning as 202 00:12:15,520 --> 00:12:15,820 a truck. 203 00:12:15,820 --> 00:12:18,730 So this is a false positive okay. 204 00:12:18,730 --> 00:12:23,260 So like in this way you can test on the video as well. 205 00:12:23,260 --> 00:12:26,650 So if I just press control C so this will interrupt this process. 206 00:12:26,650 --> 00:12:29,920 So now we have done object detection on image. 207 00:12:29,920 --> 00:12:31,720 We have done object detection on video. 208 00:12:31,720 --> 00:12:36,430 Let's move on and see how we can do object tracking using the Sort algorithm. 209 00:12:36,430 --> 00:12:41,230 So if you just go over here and if you just uh right on Google. 210 00:12:41,230 --> 00:12:42,850 So if let me just show you. 211 00:12:43,550 --> 00:12:44,960 So on GitHub. 212 00:12:45,380 --> 00:12:47,630 So you just need to open the. 213 00:12:48,630 --> 00:12:49,290 This first link. 214 00:12:49,290 --> 00:12:55,500 So sort stand for simple online real time tracking of multiple objects in a video sequence. 215 00:12:55,680 --> 00:12:58,260 Okay, so you just need this sort.py file. 216 00:13:00,050 --> 00:13:02,150 So here we have the saw.py file. 217 00:13:02,150 --> 00:13:05,360 So if you just can download this this file will be downloaded. 218 00:13:05,360 --> 00:13:08,750 But I have made some changes into the saw.py file. 219 00:13:08,930 --> 00:13:15,050 So if you use this by default saw.py board you will be in the output of. 220 00:13:15,080 --> 00:13:20,960 If you use this saw.py board in the output, you will be getting the bounding box coordinates and the 221 00:13:20,960 --> 00:13:22,940 unique IDs for each of the object. 222 00:13:22,940 --> 00:13:25,700 But I have upgraded this saw.py file code. 223 00:13:25,700 --> 00:13:31,400 And now in the output you will be getting the uh bounding box coordinates unique IDs. 224 00:13:31,400 --> 00:13:33,530 Plus you will also be getting the class names. 225 00:13:33,530 --> 00:13:38,930 So if you use this uh saw.py code over here, you will not be able to get the class names, but I have 226 00:13:38,930 --> 00:13:39,860 updated this code. 227 00:13:39,860 --> 00:13:45,230 Now you will be getting the, uh, classes name unique IDs, because in object tracking we assign a 228 00:13:45,230 --> 00:13:51,110 unique ID to each of the object, and plus you will be also be getting the bounding box coordinates. 229 00:13:51,230 --> 00:13:51,590 Okay. 230 00:13:51,590 --> 00:13:52,400 So it's your choice. 231 00:13:52,400 --> 00:13:54,680 Either you use this, but I have updated this code a bit. 232 00:13:54,680 --> 00:13:56,240 So I will be using that code. 233 00:13:56,750 --> 00:13:58,910 So let me add the code over here. 234 00:14:07,210 --> 00:14:08,890 So over here. 235 00:14:09,890 --> 00:14:10,340 Okay. 236 00:14:12,020 --> 00:14:12,770 So who are you? 237 00:14:12,770 --> 00:14:13,160 Here? 238 00:14:13,160 --> 00:14:17,750 You can see we have the updated source code and we will be using this. 239 00:14:17,840 --> 00:14:22,460 Um, we will be using this function sort over this class over here. 240 00:14:22,460 --> 00:14:25,820 And we will be using the update function over here. 241 00:14:26,000 --> 00:14:26,390 Okay. 242 00:14:26,390 --> 00:14:30,380 And from here we will be getting the bounding box coordinates, unique IDs for each of the object. 243 00:14:30,380 --> 00:14:34,100 And along with this we will also be getting the class IDs or the class names. 244 00:14:34,490 --> 00:14:37,010 Okay so let's go ahead with this. 245 00:14:37,010 --> 00:14:41,090 I am creating a new file detect-dual-tracking.py file. 246 00:14:41,090 --> 00:14:43,520 And let me add that file over here. 247 00:14:46,660 --> 00:14:50,890 So I have just created a new file detect-dual-tracking.py. 248 00:14:51,100 --> 00:14:56,200 And over here I will be writing the code to do object tracking using sort algorithm. 249 00:14:56,200 --> 00:14:56,770 The object. 250 00:14:56,770 --> 00:15:00,100 The code for object detection using yolo v9 is already added. 251 00:15:00,100 --> 00:15:06,340 I just need to add or integrate the object tracking code into this.py file, and we will be using the 252 00:15:06,370 --> 00:15:08,860 doing the object tracking using sort algorithm. 253 00:15:08,980 --> 00:15:13,510 Uh, in object tracking, what we do in object tracking, we basically assign a unique ID to each of 254 00:15:13,510 --> 00:15:14,260 the objects. 255 00:15:14,260 --> 00:15:17,080 So here you can see we have the sort.py file. 256 00:15:17,080 --> 00:15:21,490 And from the sort.py file I will be importing all the functions. 257 00:15:21,490 --> 00:15:24,070 And to import all the functions we will just write star. 258 00:15:24,070 --> 00:15:27,460 So it will import all the functions from the sort.py file. 259 00:15:29,230 --> 00:15:31,900 Next I will be just writing tracker. 260 00:15:34,470 --> 00:15:38,910 So if you just open this up for.py file over here. 261 00:15:39,330 --> 00:15:44,130 So I will be using this class uh which is this. 262 00:15:44,130 --> 00:15:47,400 So I will be using this class sort over here. 263 00:15:47,400 --> 00:15:49,710 So I will just write sort. 264 00:15:51,340 --> 00:15:53,200 And if I just see. 265 00:15:53,200 --> 00:15:55,210 So here we have the maximum age. 266 00:15:55,210 --> 00:15:57,700 So in maximum age or what we define. 267 00:15:57,700 --> 00:16:03,460 So for example if an object is lost or is there if there is any occlusion if the object is facing. 268 00:16:03,460 --> 00:16:10,690 So if for example, in simple words, if an object is lost, how many frames our sort algorithm will 269 00:16:10,690 --> 00:16:13,510 wait before assigning it a new ID. 270 00:16:13,510 --> 00:16:20,380 So for example, if an object is lost in the frame of a video, how many frames the sort algorithm will 271 00:16:20,380 --> 00:16:23,740 wait before assigning a new unique ID. 272 00:16:23,890 --> 00:16:27,640 So if I just write this value 20 over here. 273 00:16:27,640 --> 00:16:31,480 So our saw if for example, if I just write 20 over here. 274 00:16:31,480 --> 00:16:40,540 So if the car is lost in the video frame, our sort algorithm will wait 20 for 20 frames before assigning 275 00:16:40,540 --> 00:16:41,680 it a new ID. 276 00:16:41,770 --> 00:16:48,040 So maximum age we define if an object is lost in a video frame. 277 00:16:48,040 --> 00:16:54,550 How many, uh, frames our sort algorithm will wait before assigning it a new unique ID. 278 00:16:54,550 --> 00:17:02,440 So if I just write 20 over here so our sort algorithm will wait 20 frames before assigning, uh, the 279 00:17:02,440 --> 00:17:04,690 object object a new UK unique ID. 280 00:17:04,930 --> 00:17:05,560 Okay. 281 00:17:06,040 --> 00:17:06,790 So. 282 00:17:09,520 --> 00:17:16,690 So our sort algorithm will be waiting 2420 frames before assigning, uh, that object a new unique ID 283 00:17:16,690 --> 00:17:19,600 if the that object ID is lost. 284 00:17:20,680 --> 00:17:25,210 Then we will define minimum hits as three and IOU threshold. 285 00:17:25,240 --> 00:17:26,350 I have already defined this. 286 00:17:26,350 --> 00:17:27,910 What is IOU threshold at the start. 287 00:17:27,910 --> 00:17:29,920 So we'll just keep this value 0.3. 288 00:17:32,230 --> 00:17:41,980 Okay, so next what we will do is, uh, I will just go down below and, uh, where I will start writing 289 00:17:41,980 --> 00:17:43,120 the code. 290 00:17:45,230 --> 00:17:46,820 So I will just write the code. 291 00:17:47,210 --> 00:17:50,570 How we can do object tracking using sort algorithm over here. 292 00:17:58,760 --> 00:18:00,290 So let's get started with it. 293 00:18:00,290 --> 00:18:06,200 So first of all I will just create an empty numpy array of size. 294 00:18:07,180 --> 00:18:08,500 Zero six. 295 00:18:11,970 --> 00:18:18,750 Okay, so from here from the Yolo v9 algorithm, we will be getting the bounding box coordinates confidence 296 00:18:18,750 --> 00:18:20,220 score and the class name. 297 00:18:20,220 --> 00:18:23,820 So here from here is the output of the Yolo v9 algorithm. 298 00:18:23,820 --> 00:18:25,620 So if I just show you over here. 299 00:18:25,620 --> 00:18:32,190 So from the Yolo v9 algorithm we get the uh bounding box coordinates class name and confidence score. 300 00:18:32,190 --> 00:18:38,400 So we get the uh top left corner bounding box coordinates which are x one, y one. 301 00:18:38,400 --> 00:18:43,680 And for each of the detected object we get the top left corner bounding box coordinates x one, y one, 302 00:18:43,680 --> 00:18:47,970 and the bottom right corner bounding box coordinates which are x two, y two. 303 00:18:47,970 --> 00:18:51,270 So and plus we also get the class name and confidence score. 304 00:18:51,270 --> 00:18:56,310 So for each of the detected object we will be getting the top left corner bounding box coordinates X1Y1 305 00:18:56,310 --> 00:19:00,300 and bottom right corner uh bounding box coordinates X2Y2. 306 00:19:00,300 --> 00:19:02,160 So for each of the detected object. 307 00:19:08,720 --> 00:19:11,840 So over here I will just write for x, y x, y. 308 00:19:12,940 --> 00:19:16,360 So these are this is the X1Y2X1Y1. 309 00:19:16,360 --> 00:19:18,310 And this is X2Y2. 310 00:19:18,340 --> 00:19:18,880 Okay. 311 00:19:21,070 --> 00:19:23,980 So from your lovin algorithm, we get the. 312 00:19:25,040 --> 00:19:25,580 Boundingbox. 313 00:19:25,580 --> 00:19:27,560 Coordinate confidence score and class name. 314 00:19:28,970 --> 00:19:32,480 And we weren't getting this from infections. 315 00:19:32,810 --> 00:19:33,080 Okay. 316 00:19:35,630 --> 00:19:40,460 So let's separate the x1, y1 x2 y2 values. 317 00:19:45,720 --> 00:19:49,170 So from here the output is basically in the form of tensors. 318 00:19:49,170 --> 00:19:51,750 We need to convert the output from tensors into integers. 319 00:20:10,300 --> 00:20:13,690 So now we are just converting the output from tensors into integers. 320 00:20:15,060 --> 00:20:17,850 Then we just need to calculate the confidence score. 321 00:20:18,180 --> 00:20:21,330 So we just need to have math dot ceil. 322 00:20:21,330 --> 00:20:25,680 And we need to convert the confidence value into integers as well. 323 00:20:28,340 --> 00:20:31,460 You can call this as confidence score over here as well. 324 00:20:37,160 --> 00:20:40,730 Last will also be getting the class name over here. 325 00:20:43,560 --> 00:20:44,010 In. 326 00:20:47,860 --> 00:20:48,910 So. 327 00:20:48,910 --> 00:20:53,050 So these are the, uh, output which gets which we get from the Yolo v9 algorithm. 328 00:20:53,050 --> 00:20:57,610 Yolo v9 algorithm gives us the bounding box coordinates, confidence score and the class name. 329 00:20:57,610 --> 00:21:01,750 And now we'll create an array which we call grant array. 330 00:21:08,840 --> 00:21:10,580 And in this area we will have the. 331 00:21:12,510 --> 00:21:13,470 Bounding box coordinates. 332 00:21:13,470 --> 00:21:14,220 Values. 333 00:21:15,120 --> 00:21:19,860 And we'll also have the confidence score over here. 334 00:21:19,860 --> 00:21:21,420 And the class name. 335 00:21:23,690 --> 00:21:24,350 Over here. 336 00:21:24,680 --> 00:21:30,530 Okay, so let me show you how this looks like if I just print this out over here. 337 00:21:36,140 --> 00:21:37,730 Uh, I think so. 338 00:21:39,480 --> 00:21:42,540 We don't need to view the output image currently. 339 00:21:43,530 --> 00:21:44,520 And. 340 00:21:47,920 --> 00:21:55,600 Right in Thai restaurant on this other video and let me show you how this looks up. 341 00:21:56,050 --> 00:22:00,190 So in any case, if there are any error appears, we'll try to sort this up as well. 342 00:22:08,140 --> 00:22:14,170 Okay, so to run the sort algorithm skills step, we need to have the filter pi package installed as 343 00:22:14,170 --> 00:22:14,650 well. 344 00:22:14,650 --> 00:22:21,220 So I will just uh pi pip install filter pi. 345 00:22:22,150 --> 00:22:26,080 So I will just right add this up in this as well. 346 00:22:26,800 --> 00:22:27,370 Filter pi. 347 00:22:27,370 --> 00:22:30,820 So we just need to install this package individually as well. 348 00:22:31,150 --> 00:22:33,010 So let's run this up now. 349 00:22:33,010 --> 00:22:34,960 Hopefully it will run now. 350 00:22:46,370 --> 00:22:47,570 Though it has started. 351 00:22:53,250 --> 00:22:53,790 We will not. 352 00:22:53,790 --> 00:22:56,250 So now you can see over here okay. 353 00:22:56,250 --> 00:22:57,390 So okay. 354 00:22:58,340 --> 00:23:02,930 So here you can see that, uh, so for example. 355 00:23:04,880 --> 00:23:09,530 In the first frame we have 28 persons detected and there is one handbag. 356 00:23:09,980 --> 00:23:15,350 So in the first frame we have detected 28 persons and there is one handbag. 357 00:23:15,380 --> 00:23:16,100 Okay. 358 00:23:16,100 --> 00:23:20,120 So now you can see this is 26 ID which is for handbag. 359 00:23:20,570 --> 00:23:25,340 So these are the x1 y1 x2, y2 bounding box coordinates. 360 00:23:25,340 --> 00:23:26,480 This is the confidence score. 361 00:23:26,480 --> 00:23:27,920 And this is the class name. 362 00:23:28,010 --> 00:23:28,370 Okay. 363 00:23:28,370 --> 00:23:34,850 So in the same way we have already written over here like we have the bounding box coordinates and confidence 364 00:23:34,850 --> 00:23:35,960 score and class name. 365 00:23:35,960 --> 00:23:40,670 So like you can see we have detected 28 persons and one handbag. 366 00:23:40,700 --> 00:23:44,750 So for all the 28 persons we have this coordinate. 367 00:23:44,750 --> 00:23:50,240 So for each of the detected object like for the first detected object we have this bounding box coordinates. 368 00:23:50,240 --> 00:23:51,890 And we have this confidence score. 369 00:23:51,890 --> 00:23:57,530 And this is the class name for the second object uh detected object we have this bounding box coordinates. 370 00:23:57,620 --> 00:23:59,510 And this is the confidence score. 371 00:23:59,510 --> 00:24:02,360 And this is the class name for the third detected object. 372 00:24:02,360 --> 00:24:03,830 These are the bounding box coordinates. 373 00:24:03,830 --> 00:24:05,270 This is the confidence score. 374 00:24:05,270 --> 00:24:06,350 And this is the class name. 375 00:24:06,350 --> 00:24:10,970 So we are just getting the bounding box coordinates confidence score and class name for each of the 376 00:24:10,970 --> 00:24:12,830 object one by one over here. 377 00:24:15,030 --> 00:24:18,270 So let's go ahead with this up as well now. 378 00:24:21,060 --> 00:24:23,940 So now I will just write detection is equal to NP dot p stack. 379 00:24:23,940 --> 00:24:26,010 So we will be doing vertical stack. 380 00:24:33,070 --> 00:24:34,150 Deductions. 381 00:24:37,140 --> 00:24:37,680 Country. 382 00:24:38,220 --> 00:24:41,040 So now let me show you what this vertical stack does. 383 00:24:41,040 --> 00:24:42,660 So if I just write. 384 00:24:43,880 --> 00:24:44,330 Then. 385 00:24:47,200 --> 00:24:48,370 Actions. 386 00:24:50,380 --> 00:24:50,770 Okay. 387 00:24:52,020 --> 00:24:54,630 So let's run this script again. 388 00:25:17,560 --> 00:25:18,220 So let me show you. 389 00:25:18,220 --> 00:25:19,780 We have the for the first frame. 390 00:25:20,850 --> 00:25:21,840 Okay. 391 00:25:24,120 --> 00:25:26,610 So let me just roll it up now. 392 00:25:30,440 --> 00:25:34,940 So now you can see that these are the bounding box coordinates for the first detection. 393 00:25:35,330 --> 00:25:35,600 Okay. 394 00:25:35,600 --> 00:25:36,740 This is the confidence score. 395 00:25:36,740 --> 00:25:38,360 And this is the class name. 396 00:25:38,510 --> 00:25:39,050 Okay. 397 00:25:39,050 --> 00:25:43,610 So now like you can see that this is the first uh array. 398 00:25:43,850 --> 00:25:44,360 Okay. 399 00:25:44,360 --> 00:25:50,750 And then for the so now you can see that these are the bounding box coordinates for the second detection. 400 00:25:50,750 --> 00:25:52,670 And this is the confidence score. 401 00:25:52,670 --> 00:25:53,960 And this is the class name. 402 00:25:53,960 --> 00:26:00,170 But you can see over here we have also added the first detected object bounding box coordinates. 403 00:26:01,210 --> 00:26:05,440 Contnued confidence score and class name into this numpy array as well. 404 00:26:05,470 --> 00:26:07,540 Now if I just code over here. 405 00:26:07,540 --> 00:26:12,730 Now these are the bounding box coordinates confidence score and class name for the first two detections. 406 00:26:12,730 --> 00:26:14,770 Now here you can see we have for the third detection. 407 00:26:14,770 --> 00:26:19,390 And I have added the bounding box coordinates confidence score and class name for the first two detections 408 00:26:19,390 --> 00:26:19,930 as well. 409 00:26:19,930 --> 00:26:22,330 So we are doing vertical stack over here. 410 00:26:22,570 --> 00:26:27,670 And here you can see these are the bounding box coordinates confidence score and the class name for 411 00:26:27,670 --> 00:26:28,780 the for detection. 412 00:26:28,780 --> 00:26:33,820 But I have added the bounding box coordinate confidence score and class name for the first three detections 413 00:26:33,820 --> 00:26:35,500 as well okay. 414 00:26:37,110 --> 00:26:39,270 So this is how it works. 415 00:26:49,020 --> 00:26:50,730 Now I will just write. 416 00:26:53,880 --> 00:26:54,390 Tracker. 417 00:26:54,390 --> 00:27:00,360 So you can see that I've initialized the sort, uh, algorithm over above over here. 418 00:27:00,360 --> 00:27:01,830 So we initialized over here. 419 00:27:01,830 --> 00:27:03,000 So we have tracker. 420 00:27:04,320 --> 00:27:08,520 Okay, so now I will just write over here tracker. 421 00:27:08,520 --> 00:27:12,120 So I will be using the update function inside this sort. 422 00:27:12,150 --> 00:27:13,110 Dot py file. 423 00:27:13,110 --> 00:27:18,450 So I'm using the class sort and I will be using this update function over here which you can see. 424 00:27:18,600 --> 00:27:19,500 So. 425 00:27:21,610 --> 00:27:25,690 Okay, so let's use the tracker dot update. 426 00:27:28,140 --> 00:27:30,240 And what this function accepts. 427 00:27:30,240 --> 00:27:32,070 So this function accepts reductions. 428 00:27:32,070 --> 00:27:36,060 And over here I will passing this reductions over here. 429 00:27:38,010 --> 00:27:38,550 Okay. 430 00:27:38,550 --> 00:27:43,860 So so now before we go ahead, uh, the, the sort algorithm. 431 00:27:43,860 --> 00:27:49,410 So the input of the Sort algorithm, we have passed out the bounding box coordinates confidence score 432 00:27:49,410 --> 00:27:50,220 and class name. 433 00:27:50,220 --> 00:27:54,030 So what output we get from the sort algorithm okay. 434 00:27:54,690 --> 00:27:56,730 So if I just open this from here. 435 00:27:56,730 --> 00:28:02,250 So from the sort object tracking algorithm we get out we get the bounding box coordinates identities, 436 00:28:02,250 --> 00:28:04,680 unique IDs for each of the object and class name. 437 00:28:04,680 --> 00:28:09,060 So we don't get the confidence score from the sort object tracking algorithm output. 438 00:28:09,060 --> 00:28:11,910 We get the unique IDs for each of the detected object. 439 00:28:12,730 --> 00:28:13,030 Okay. 440 00:28:16,690 --> 00:28:17,920 So I will just write. 441 00:28:19,370 --> 00:28:22,010 So if there are any detections. 442 00:28:26,940 --> 00:28:29,400 So this will return us the bounding box coordinates. 443 00:28:35,140 --> 00:28:38,350 So if you just print this tracker detections. 444 00:28:38,350 --> 00:28:41,920 Uh, let me just print this out first and then we will discuss. 445 00:28:48,280 --> 00:28:49,030 Okay? 446 00:28:49,540 --> 00:28:50,050 Yeah, sure. 447 00:28:51,220 --> 00:28:51,580 Soon. 448 00:28:56,740 --> 00:28:58,570 Well, let's run this up first. 449 00:29:09,440 --> 00:29:12,440 So this will take a few seconds before it starts executing. 450 00:29:22,040 --> 00:29:23,600 Okay, so here we have it. 451 00:29:24,460 --> 00:29:30,490 So if you just see over here so the first four have the bounding box coordinates. 452 00:29:30,820 --> 00:29:31,090 Okay. 453 00:29:31,090 --> 00:29:33,070 So if I just write over here. 454 00:29:35,590 --> 00:29:36,040 So. 455 00:29:41,250 --> 00:29:43,800 In the first four, we have the bounding box coordinates. 456 00:29:50,910 --> 00:29:57,270 And these are 0012345678. 457 00:29:57,270 --> 00:30:00,870 The eight item contain the unique IDs. 458 00:30:04,740 --> 00:30:05,400 Okay. 459 00:30:05,970 --> 00:30:07,200 And. 460 00:30:09,130 --> 00:30:14,440 01234, and this fourth item contains the class names. 461 00:30:14,440 --> 00:30:15,040 Okay. 462 00:30:17,300 --> 00:30:19,550 Or these are the class IDs you can say. 463 00:30:21,230 --> 00:30:24,860 So we can simply write get grease. 464 00:30:31,030 --> 00:30:35,890 So this four item, uh, 01234 contain the class IDs. 465 00:30:38,290 --> 00:30:44,020 So now I have just created a function over here by the name draw those boxes inside the draw boxes. 466 00:30:44,020 --> 00:30:44,740 Function. 467 00:30:44,920 --> 00:30:45,340 Function. 468 00:30:45,340 --> 00:30:46,390 Draw boxes function. 469 00:30:46,390 --> 00:30:50,320 We pass down the bounding box coordinates and entities and the class ids. 470 00:30:50,320 --> 00:30:53,890 For example, the class ID for the person class is zero. 471 00:30:53,890 --> 00:30:55,330 For the bicycle it's one. 472 00:30:55,720 --> 00:30:56,260 Okay. 473 00:30:56,260 --> 00:30:58,210 And this, uh, what? 474 00:30:58,210 --> 00:31:01,840 In the draw box function I've just written using OpenCV. 475 00:31:02,020 --> 00:31:06,100 Uh cv2 dot rectangle function OpenCV cv2 dot rectangle function. 476 00:31:06,100 --> 00:31:10,000 We are drawing the bounding box coordinates or rectangles around each of the detected object. 477 00:31:10,000 --> 00:31:15,340 Plus we are also creating a rectangle above the detected object, where we will be adding the IDs and 478 00:31:15,340 --> 00:31:17,950 the class names over there as well. 479 00:31:17,950 --> 00:31:24,370 So in the label we will have the id and the class name, and we will be drawing a circle at the center 480 00:31:24,370 --> 00:31:26,290 of each of the detected object as well. 481 00:31:26,290 --> 00:31:28,960 And let me discuss this in detail okay. 482 00:31:28,960 --> 00:31:32,980 So I will just calling the draw boxes function over here. 483 00:31:34,110 --> 00:31:36,270 Well, this is the input video frame. 484 00:31:38,280 --> 00:31:38,790 Morning. 485 00:31:38,790 --> 00:31:39,780 Most gardeners. 486 00:31:41,010 --> 00:31:44,190 Unique IDs for each of the detected object and the categories. 487 00:31:44,250 --> 00:31:45,900 Okay, so. 488 00:31:47,170 --> 00:31:50,350 So now you can see I have this path to the draw boxes function. 489 00:31:50,350 --> 00:31:52,300 And here we have this function. 490 00:31:52,300 --> 00:31:55,870 So here we are just finding the center point of the bounding box. 491 00:31:55,870 --> 00:32:03,430 We are just uh have finding the unique I and uh, giving the unique IDs to each of the object which 492 00:32:03,430 --> 00:32:07,600 we have or which we are already getting the sort algorithm, we are drawing drawing rectangle around 493 00:32:07,600 --> 00:32:08,620 each of the detected object. 494 00:32:08,620 --> 00:32:11,080 Plus we have just created this function class names. 495 00:32:11,080 --> 00:32:16,120 So if we get the object id zero this mean it represent the detected object is the person. 496 00:32:16,120 --> 00:32:22,630 If the get uh class ID or categories, which we have called this as one, it is a bicycle restricted. 497 00:32:22,630 --> 00:32:24,820 If the class ID which we got is two. 498 00:32:24,850 --> 00:32:26,170 So the car is detected. 499 00:32:26,170 --> 00:32:28,750 If the class ID is three then the motorbike is detected. 500 00:32:28,750 --> 00:32:31,480 If the car class ID is four then the aeroplane is detected. 501 00:32:31,480 --> 00:32:33,820 If the class ID is five, then the bus is detected. 502 00:32:33,820 --> 00:32:36,820 If the class ID is 80, then the toothbrush is detected. 503 00:32:36,820 --> 00:32:40,960 If the class ID is, uh 79, then here I draw restricted. 504 00:32:40,960 --> 00:32:41,470 Okay. 505 00:32:41,470 --> 00:32:42,280 So. 506 00:32:43,520 --> 00:32:45,110 You can see over here. 507 00:32:45,440 --> 00:32:48,770 Um, so here we have the class IDs. 508 00:32:48,770 --> 00:32:51,410 So, so in the categories we have the class IDs. 509 00:32:51,410 --> 00:32:55,610 And this ID represent a unique ID which we assign to each of the object. 510 00:32:55,640 --> 00:32:56,030 Okay. 511 00:32:56,030 --> 00:32:59,570 So in object tracking we basically assign a unique ID to each of the object. 512 00:32:59,570 --> 00:33:04,040 So you can see that in this cat we cat cat we have the class ID. 513 00:33:04,040 --> 00:33:07,460 And when this we pass the class ID over here. 514 00:33:07,460 --> 00:33:11,030 So if the class ID is zero then we will be calling this function over here. 515 00:33:11,030 --> 00:33:14,030 And it will tell us that this is a person class and so on. 516 00:33:14,090 --> 00:33:14,720 Okay. 517 00:33:14,720 --> 00:33:19,760 And in the label we have the unique ID that we have assigned to each of the object plus the class name 518 00:33:19,760 --> 00:33:20,510 over here. 519 00:33:22,200 --> 00:33:26,970 And then we are just drawing a circle around each of us at the center of each of the detected object. 520 00:33:27,480 --> 00:33:31,800 Okay, so now I will just run this script over here. 521 00:33:32,980 --> 00:33:39,190 So plus I will just write view dash image and let's run this up over here. 522 00:33:40,950 --> 00:33:46,530 Let's see if we are able to detect the objects and assign a unique ID to each of the detected object 523 00:33:46,530 --> 00:33:48,540 using the Sort algorithm or not. 524 00:33:57,320 --> 00:33:58,580 So over here. 525 00:33:58,580 --> 00:33:59,240 That's great. 526 00:33:59,240 --> 00:34:06,290 So now you can see over here, uh, each detected object is being assigned a unique ID six 232 2314 527 00:34:06,290 --> 00:34:15,620 1710 2925 2119 eight 1813 1516, 1231 2020. 528 00:34:15,620 --> 00:34:20,180 So you can see that using object tracking algorithm, we are able to assign a unique ID to each of the 529 00:34:20,180 --> 00:34:20,960 detected object. 530 00:34:20,960 --> 00:34:22,340 And that's amazing. 531 00:34:22,340 --> 00:34:23,210 That's great. 532 00:34:23,510 --> 00:34:26,300 Okay so I'm testing on CPU. 533 00:34:26,300 --> 00:34:27,920 So the processing is a bit slow. 534 00:34:27,920 --> 00:34:31,340 Let me just test it on other video as well. 535 00:34:56,130 --> 00:35:01,020 So now you can see over here we have assigned a unique ID to the car as well. 536 00:35:01,020 --> 00:35:04,800 So this is the wrong detection but a unique ID assigned to each of the object. 537 00:35:04,800 --> 00:35:09,930 Because we assign a unique ID to each of the detections using sort algorithm. 538 00:35:10,230 --> 00:35:10,620 Okay. 539 00:35:10,620 --> 00:35:12,900 So one for two. 540 00:35:13,050 --> 00:35:15,900 So that will look pretty fine as well. 541 00:35:16,080 --> 00:35:19,500 Like you can see let's see if we are able to detect this car or not. 542 00:35:19,500 --> 00:35:24,030 So you can see the car is detected and unique ID seven is assigned over here as well. 543 00:35:24,030 --> 00:35:25,770 And this works pretty good. 544 00:35:26,800 --> 00:35:32,350 So if you want to test it on the live webcam feed, you can, uh, simply right over here in the source 545 00:35:32,350 --> 00:35:32,800 zero. 546 00:35:32,800 --> 00:35:36,550 So if you write source zero, you will be able to test on the live webcam feed. 547 00:35:36,550 --> 00:35:39,670 And if you want to test on the external webcam, you can write one. 548 00:35:39,670 --> 00:35:43,720 And if you want to test, uh, test on the uh, like IP camera. 549 00:35:43,720 --> 00:35:47,050 So you can just pass the IP, um, link over here. 550 00:35:47,050 --> 00:35:50,620 So if you want to test on the, uh, your laptop webcam, you can just write zero. 551 00:35:50,620 --> 00:35:53,950 If you want to test on external webcam, you can write one over here. 552 00:35:53,950 --> 00:36:00,670 So if you want to test, uh, if you want to detect only specific classes, like if you just write over 553 00:36:00,670 --> 00:36:03,280 here like test dot mp4. 554 00:36:03,280 --> 00:36:08,770 So if you want to like you want to only detect uh person class okay. 555 00:36:08,770 --> 00:36:12,130 So the person class ID is. 556 00:36:14,510 --> 00:36:15,260 Zero. 557 00:36:15,470 --> 00:36:21,380 Okay, so you can see over here, uh, if I just show you the person class ID zero. 558 00:36:21,380 --> 00:36:22,430 So I will just write. 559 00:36:22,430 --> 00:36:23,210 Class is zero. 560 00:36:23,210 --> 00:36:27,140 So this will only detect the person class and no other object will be detected. 561 00:36:27,140 --> 00:36:28,850 So you can do this as well. 562 00:36:35,470 --> 00:36:38,980 Oh, this will take few seconds before this script gets executed. 563 00:36:46,650 --> 00:36:49,380 Uh, one thing we can do is. 564 00:36:51,410 --> 00:36:53,150 You can simply remove the sperm. 565 00:36:55,890 --> 00:36:58,620 So let's see how our output looks like. 566 00:37:00,550 --> 00:37:04,660 Okay, so now you can see over here, uh, we are only detecting the person class. 567 00:37:04,660 --> 00:37:05,830 No hand-bags. 568 00:37:05,830 --> 00:37:08,560 No other thing is, uh, detected over here. 569 00:37:08,560 --> 00:37:13,420 Okay, so now you can see that we are only detecting the person class. 570 00:37:13,420 --> 00:37:16,480 No hand-bags or no other things are being detected. 571 00:37:16,480 --> 00:37:22,420 So you can select any specific class which you want to detect in a video frame feed as well, or in 572 00:37:22,420 --> 00:37:23,620 the live webcam feed. 573 00:37:23,620 --> 00:37:25,570 So that's all from this tutorial. 574 00:37:25,570 --> 00:37:26,440 Thank you for watching! 575 00:37:26,440 --> 00:37:31,060 In this tutorial, we have learned how we can integrate object detection with object tracking. 576 00:37:31,060 --> 00:37:34,840 And here we are using, uh, sort algorithm to do object tracking. 577 00:37:34,840 --> 00:37:37,840 And for the object detection we are using the Yolo v nine algorithm. 578 00:37:37,840 --> 00:37:38,590 Thank you for watching. 579 00:37:38,590 --> 00:37:38,950 Bye bye.