1 00:00:03,000 --> 00:00:09,000 In this video tutorial, we will learn how we can create a Streamlit application to do object detection, 2 00:00:09,000 --> 00:00:13,000 instance segmentation and pose estimation using Ultralytics. 3 00:00:13,000 --> 00:00:14,000 YOLO 11. 4 00:00:14,000 --> 00:00:20,000 Here is a quick demo of the Streamlit application that we will be creating in this tutorial. 5 00:00:20,000 --> 00:00:26,000 So let me first give you a quick demo of the Streamlit application that we are building in this tutorial. 6 00:00:26,000 --> 00:00:32,000 So we can perform object detection, instance segmentation and pose estimation with the help of this 7 00:00:32,000 --> 00:00:33,000 Streamlit application. 8 00:00:33,000 --> 00:00:40,000 And we will be able to perform object detection, instance segmentation and pose estimation on images 9 00:00:40,000 --> 00:00:41,000 and on videos. 10 00:00:41,000 --> 00:00:43,000 You can upload any random image like. 11 00:00:43,000 --> 00:00:48,000 You can click over here from Raw files and you can upload any image from here. 12 00:00:49,000 --> 00:00:53,000 Okay, as you upload the image, an image will appear over here. 13 00:00:53,000 --> 00:01:00,000 And if you want to perform a object detection, you can see over here we have by default the object 14 00:01:00,000 --> 00:01:02,000 detection is being marked over here. 15 00:01:02,000 --> 00:01:04,000 And then you can click on Detect object. 16 00:01:04,000 --> 00:01:07,000 And over here you will see an output image. 17 00:01:07,000 --> 00:01:09,000 Once the object detection is being performed. 18 00:01:10,000 --> 00:01:13,000 So now you can see over here we have the output image. 19 00:01:13,000 --> 00:01:19,000 And you can see we are able to detect bus person uh backpack handbags. 20 00:01:19,000 --> 00:01:24,000 So you can see the detection results over here like you can see now uh and over here you can see we 21 00:01:24,000 --> 00:01:29,000 have the, uh, bounding box coordinates over here in the form of tensors. 22 00:01:29,000 --> 00:01:35,000 Okay, so we have the bounding box coordinates for each of the detected object in the form of tensors. 23 00:01:35,000 --> 00:01:35,000 Okay. 24 00:01:35,000 --> 00:01:39,000 So as you expand this, you will see all the bounding box coordinates. 25 00:01:39,000 --> 00:01:45,000 And here you can see over here we have are able to detect a person backpack and bag person or bus and 26 00:01:45,000 --> 00:01:46,000 all the things. 27 00:01:46,000 --> 00:01:49,000 Okay so currently I'm using YOLO 11 nano model. 28 00:01:49,000 --> 00:01:55,000 And if you use uh large models of YOLO 11 you will get definitely better results. 29 00:01:56,000 --> 00:02:00,000 So going ahead you can also perform a segmentation as well. 30 00:02:00,000 --> 00:02:06,000 So if you want to perform the segmentation on the same image, you can click on segmentation from here. 31 00:02:06,000 --> 00:02:09,000 And you can just go down from here and click on Detect Object. 32 00:02:10,000 --> 00:02:11,000 Okay. 33 00:02:13,000 --> 00:02:16,000 So now you can see over here we are able to perform the segmentation. 34 00:02:16,000 --> 00:02:19,000 We are able to take the person backpack. 35 00:02:19,000 --> 00:02:21,000 And we are able to draw the mask as well. 36 00:02:21,000 --> 00:02:27,000 So basically in instance segmentation we can draw a mask around each of the detected object. 37 00:02:27,000 --> 00:02:32,000 And you can see over here we have we are able to draw a blue mask around each of the detected object 38 00:02:32,000 --> 00:02:33,000 okay. 39 00:02:33,000 --> 00:02:36,000 In the same way you can also perform pose estimation. 40 00:02:36,000 --> 00:02:41,000 And if you click on pose estimation from here and going below, if you click on Detect object from here. 41 00:02:45,000 --> 00:02:48,000 So now you can see over here we are able to perform the pose estimation. 42 00:02:48,000 --> 00:02:53,000 And you can see that we have drawn the key points around each of the detected object. 43 00:02:53,000 --> 00:02:57,000 Like you can see over here we are able to perform pose estimation. 44 00:02:57,000 --> 00:03:01,000 And we are able to draw or draw the key points around each of the detected object. 45 00:03:02,000 --> 00:03:02,000 Okay. 46 00:03:02,000 --> 00:03:09,000 In the same way, if you want to upload some other image, you can just click on upload files from here 47 00:03:09,000 --> 00:03:11,000 and you can upload this image from here. 48 00:03:15,000 --> 00:03:16,000 Okay. 49 00:03:16,000 --> 00:03:22,000 And after uploading this image, if you want to perform pose estimation, I have already selected pose 50 00:03:22,000 --> 00:03:22,000 estimation. 51 00:03:22,000 --> 00:03:24,000 And you can click on Detect Objects. 52 00:03:24,000 --> 00:03:27,000 So now you can see over here we are able to detect the object. 53 00:03:27,000 --> 00:03:31,000 Plus we are also able to draw the key points around each object object. 54 00:03:31,000 --> 00:03:32,000 So we have performed pose estimation. 55 00:03:32,000 --> 00:03:33,000 In the same way. 56 00:03:33,000 --> 00:03:38,000 If you want to perform segmentation you can just click on segmentation from here and you will be able 57 00:03:38,000 --> 00:03:40,000 to perform segmentation. 58 00:03:40,000 --> 00:03:45,000 Now you can see over here we are able to detect the object as well as we are able to draw a mask around 59 00:03:45,000 --> 00:03:47,000 each of the detected object over here. 60 00:03:47,000 --> 00:03:48,000 Okay. 61 00:03:48,000 --> 00:03:54,000 So we have seen that how we can perform object detection, instance segmentation and pose estimation 62 00:03:55,000 --> 00:03:56,000 on the image. 63 00:03:56,000 --> 00:03:58,000 Let's go to the video part. 64 00:03:58,000 --> 00:04:05,000 So if you want to perform object detection, instance segmentation and pose estimation on video, you 65 00:04:05,000 --> 00:04:07,000 can just select video from here as the source. 66 00:04:07,000 --> 00:04:09,000 And you can see this is the default video. 67 00:04:09,000 --> 00:04:15,000 Let's first see how we can perform object detection, instance segmentation and pose estimation on a 68 00:04:15,000 --> 00:04:16,000 video. 69 00:04:16,000 --> 00:04:19,000 So in my video directory I have added the two different videos. 70 00:04:19,000 --> 00:04:26,000 You can add as many videos you want, so as many videos you add in your video directory, you will be 71 00:04:27,000 --> 00:04:31,000 able to perform object detection, instance segmentation and pose estimation on those videos. 72 00:04:31,000 --> 00:04:37,000 Okay, so if you select video one and if you want to perform object detection like let's click on Detect 73 00:04:37,000 --> 00:04:37,000 Object. 74 00:04:37,000 --> 00:04:38,000 From here. 75 00:04:38,000 --> 00:04:40,000 And here you can see the output. 76 00:04:40,000 --> 00:04:43,000 We are able to detect the person over here okay. 77 00:04:44,000 --> 00:04:47,000 So like this is the wrong detection. 78 00:04:47,000 --> 00:04:49,000 Currently I'm using the 11 nano model. 79 00:04:49,000 --> 00:04:55,000 And if you want to get good results you can use a higher YOLO 11 models which include your 11 large. 80 00:04:55,000 --> 00:04:57,000 Your or 11 extra large. 81 00:04:57,000 --> 00:05:02,000 Uh, the reason I'm using a small yellow model, like 11 nano is that I'm currently running, uh, the 82 00:05:02,000 --> 00:05:08,000 Streamlit application on my CPU, and I don't have GPU in my local machine, so YOLO 11 is the fast. 83 00:05:08,000 --> 00:05:15,000 YOLO 11 nano is the fastest, but it is least accurate, while YOLO 11 Extra Large is the most accurate. 84 00:05:15,000 --> 00:05:17,000 But it takes more inference time. 85 00:05:17,000 --> 00:05:21,000 Okay, so there is a trade off between speed and accuracy. 86 00:05:22,000 --> 00:05:27,000 Okay, so if you want to go with another video like you can see over here, you can click on Detect 87 00:05:27,000 --> 00:05:32,000 Video Objects and you can see over here we are able to detect the person bicycle over here. 88 00:05:32,000 --> 00:05:35,000 And detection results look quite promising. 89 00:05:35,000 --> 00:05:36,000 Okay. 90 00:05:36,000 --> 00:05:41,000 And if you want to perform segmentation you can click on segmentation from here and click on Detect 91 00:05:41,000 --> 00:05:42,000 Video Objects. 92 00:05:42,000 --> 00:05:45,000 And now you can see that we are able to perform segmentation. 93 00:05:45,000 --> 00:05:48,000 We are able to draw a mask around each of the directory object. 94 00:05:49,000 --> 00:05:53,000 Uh the person are in the blue mask which you can clearly see over here. 95 00:05:53,000 --> 00:05:58,000 And if you want to perform pose estimation, you can just click on Pose Estimation and click on Detect 96 00:05:58,000 --> 00:05:59,000 Video Objects. 97 00:05:59,000 --> 00:06:02,000 And now you can see that we are able to detect the objects. 98 00:06:02,000 --> 00:06:06,000 And we are also able to draw key points around each of the object over here. 99 00:06:07,000 --> 00:06:13,000 You can also just a confidence threshold, like if you increase the confident threshold to 0.933 like 100 00:06:13,000 --> 00:06:18,000 93%, you can see detection has um like there are no detections. 101 00:06:18,000 --> 00:06:22,000 And if I just reduce the confidence to 70 let's see. 102 00:06:22,000 --> 00:06:28,000 So now you can see that, uh, it will only detect the object that have a confidence score above 70% 103 00:06:28,000 --> 00:06:30,000 or 0.7. 104 00:06:30,000 --> 00:06:35,000 So all the objects that have confidence score above 70% will be detected. 105 00:06:35,000 --> 00:06:37,000 So what does confidence score means? 106 00:06:37,000 --> 00:06:42,000 Like the confidence score means how much the model is confident that this is a person or this is a bicycle 107 00:06:42,000 --> 00:06:43,000 okay. 108 00:06:44,000 --> 00:06:48,000 So if you just increase the confidence score there will be less false positives. 109 00:06:48,000 --> 00:06:55,000 But if you decrease the confidence score like maybe to 25, or if you further increase the confidence, 110 00:06:55,000 --> 00:06:57,000 go to 10% or 0.10. 111 00:06:57,000 --> 00:07:02,000 You will get more detections, but there will be more false positives as well. 112 00:07:02,000 --> 00:07:07,000 Okay, but if you increase the uh, confidence score, there will be more accurate detection and there 113 00:07:07,000 --> 00:07:11,000 will be less false positives, but there will be few detections as well. 114 00:07:11,000 --> 00:07:12,000 Okay. 115 00:07:12,000 --> 00:07:15,000 So this is how our Streamlit application works. 116 00:07:16,000 --> 00:07:19,000 And we will be creating this Streamlit application in this tutorial. 117 00:07:19,000 --> 00:07:21,000 So let's get started with this. 118 00:07:23,000 --> 00:07:28,000 So over here you can see that I have just created a project in PyCharm Community Edition. 119 00:07:28,000 --> 00:07:31,000 And I have requirements.txt file. 120 00:07:31,000 --> 00:07:35,000 And we are we have Ultralytics and Streamlit package over here. 121 00:07:35,000 --> 00:07:39,000 So if you create this application we only require two packages. 122 00:07:39,000 --> 00:07:41,000 One is Ultralytics and Streamlit. 123 00:07:41,000 --> 00:07:45,000 And to install this is you can simply right pip install. 124 00:07:47,000 --> 00:07:52,000 R requirements dot txt. 125 00:07:52,000 --> 00:07:56,000 And if you run this you will be able to install these two packages. 126 00:07:56,000 --> 00:08:00,000 So I've already installed these packages and you can see over here requirements already satisfied. 127 00:08:00,000 --> 00:08:04,000 But if you are installing these packages for the first time this will take some time. 128 00:08:04,000 --> 00:08:08,000 Next we will just create a new file like Main.py file. 129 00:08:10,000 --> 00:08:16,000 In the step number one, we will import all the required libraries that we require in this project. 130 00:08:20,000 --> 00:08:22,000 So we require the OpenCV Python package. 131 00:08:22,000 --> 00:08:24,000 We require the Streamlit library. 132 00:08:35,000 --> 00:08:36,000 To define the path. 133 00:08:38,000 --> 00:08:42,000 Of our root directory we require the Pathlib primary. 134 00:08:42,000 --> 00:08:43,000 From there we import path. 135 00:08:45,000 --> 00:08:48,000 Then to require SSIS library. 136 00:08:50,000 --> 00:08:53,000 Then to go to object detection using Ultralytics YOLO 11. 137 00:08:54,000 --> 00:08:59,000 We require a YOLO package software report from Ultralytics. 138 00:08:59,000 --> 00:09:01,000 We will import YOLO. 139 00:09:06,000 --> 00:09:10,000 And to display any image we require image lab. 140 00:09:14,000 --> 00:09:15,000 Okay. 141 00:09:21,000 --> 00:09:27,000 So let's create a new page layout and with the help of Streamlit. 142 00:09:28,000 --> 00:09:31,000 So we will be now creating an initial page layout. 143 00:09:34,000 --> 00:09:37,000 So we'll write speed dot set dash page dash config. 144 00:09:39,000 --> 00:09:42,000 And the page title will be. 145 00:09:45,000 --> 00:09:46,000 YOLO 11. 146 00:09:47,000 --> 00:09:49,000 We can also add an page icon. 147 00:10:02,000 --> 00:10:05,000 Next we will define a header for our Streamlit application. 148 00:10:11,000 --> 00:10:15,000 The header will be object detection level. 149 00:10:18,000 --> 00:10:20,000 We can also create a sidebar. 150 00:10:20,000 --> 00:10:26,000 We will have our all the settings like what task we want to perform, what would be the confidence threshold. 151 00:10:26,000 --> 00:10:30,000 And if you want to do object detection on image or on video. 152 00:10:30,000 --> 00:10:33,000 So you need to create a sidebar for this. 153 00:10:39,000 --> 00:10:45,000 In the sidebar we will also add a header with the name model configurations. 154 00:10:53,000 --> 00:10:53,000 In the sidebar. 155 00:10:53,000 --> 00:10:58,000 We will also give the user the option to choose whether you want to perform object detection, instance 156 00:10:58,000 --> 00:11:00,000 segmentation, or pose estimation. 157 00:11:13,000 --> 00:11:16,000 Or the user will have the option to choose the model type. 158 00:11:28,000 --> 00:11:32,000 We are just adding radio buttons for their. 159 00:11:37,000 --> 00:11:38,000 Detection. 160 00:11:38,000 --> 00:11:39,000 Segmentation. 161 00:11:42,000 --> 00:11:43,000 Fault estimation. 162 00:11:48,000 --> 00:11:54,000 We will also give the user the option to select the confidence threshold with the help of slider. 163 00:12:03,000 --> 00:12:05,000 The confidence value will be in the form of float. 164 00:12:22,000 --> 00:12:24,000 The minimum value of confidence will be 25. 165 00:12:24,000 --> 00:12:28,000 The default value will be 40 and the maximum value will be 100. 166 00:12:30,000 --> 00:12:36,000 We will divide it by 100 because our model takes the confidence value in the form of, uh, between 167 00:12:36,000 --> 00:12:43,000 0 to 1, like 0.1 is 10%, 0.25 is 25%, 0.90 is 90%. 168 00:12:43,000 --> 00:12:43,000 Okay. 169 00:12:44,000 --> 00:12:45,000 So we write. 170 00:12:49,000 --> 00:12:52,000 So let's see how each Streamlit application looks like. 171 00:12:52,000 --> 00:12:56,000 So we'll write Streamlit run main.py. 172 00:13:23,000 --> 00:13:25,000 So this will take some time. 173 00:13:25,000 --> 00:13:27,000 Currently this will take some time. 174 00:13:28,000 --> 00:13:29,000 Let me see if there is an error. 175 00:13:31,000 --> 00:13:32,000 It looks fine. 176 00:13:38,000 --> 00:13:38,000 Okay. 177 00:13:38,000 --> 00:13:42,000 So here you can see we have our Streamlit application. 178 00:13:42,000 --> 00:13:46,000 Like you can see over here we have the model configurations cost reduction. 179 00:13:46,000 --> 00:13:47,000 Segmentation. 180 00:13:47,000 --> 00:13:48,000 Cost estimation. 181 00:13:48,000 --> 00:13:49,000 Okay. 182 00:13:49,000 --> 00:13:50,000 Default value. 183 00:13:50,000 --> 00:13:52,000 Why we are seeing this. 184 00:13:59,000 --> 00:13:59,000 Okay. 185 00:13:59,000 --> 00:14:03,000 So the maximum value is 100 and default value is 40. 186 00:14:05,000 --> 00:14:06,000 I have just made this news. 187 00:14:06,000 --> 00:14:07,000 Let's see. 188 00:14:08,000 --> 00:14:10,000 Let me refresh this up. 189 00:14:13,000 --> 00:14:13,000 Okay. 190 00:14:13,000 --> 00:14:14,000 Again I think. 191 00:14:18,000 --> 00:14:18,000 Oh yeah. 192 00:14:20,000 --> 00:14:21,000 Let me refresh this up. 193 00:14:24,000 --> 00:14:25,000 So you can see. 194 00:14:25,000 --> 00:14:28,000 Default value is 40 and the minimum value is 25. 195 00:14:28,000 --> 00:14:30,000 And maximum value is 100. 196 00:14:30,000 --> 00:14:30,000 Okay. 197 00:14:30,000 --> 00:14:33,000 So let's complete this application now. 198 00:14:39,000 --> 00:14:45,000 First we need to get the absolute path of the current file. 199 00:14:45,000 --> 00:14:45,000 Wide. 200 00:15:00,000 --> 00:15:07,000 And we need to get the parent directory of the parent file. 201 00:15:09,000 --> 00:15:10,000 Which is. 202 00:15:19,000 --> 00:15:22,000 Next we will add the root path. 203 00:15:25,000 --> 00:15:27,000 To the system dot. 204 00:15:28,000 --> 00:15:29,000 Or if it's not already there. 205 00:15:41,000 --> 00:15:44,000 I will just say if not in. 206 00:15:46,000 --> 00:15:48,000 This pod. 207 00:15:51,000 --> 00:15:53,000 Dot dot dot. 208 00:15:53,000 --> 00:15:54,000 Append. 209 00:15:59,000 --> 00:15:59,000 Okay. 210 00:16:04,000 --> 00:16:07,000 Next we need to get the relative path. 211 00:16:12,000 --> 00:16:14,000 Of the root directory. 212 00:16:17,000 --> 00:16:23,000 With respect to the current working directory. 213 00:16:41,000 --> 00:16:42,000 Okay. 214 00:16:42,000 --> 00:16:48,000 So we have just got that part to the root directory with respect to the current current working directory. 215 00:16:48,000 --> 00:16:50,000 Uh, that looks fine. 216 00:16:50,000 --> 00:16:51,000 Next. 217 00:16:53,000 --> 00:16:55,000 So we need to define the sources. 218 00:16:55,000 --> 00:17:03,000 Uh, our we can pass our input in the form of image and as well as in the form of video. 219 00:17:10,000 --> 00:17:13,000 Our input will be in the form of image or video. 220 00:17:16,000 --> 00:17:18,000 So let's define a source list. 221 00:17:41,000 --> 00:17:47,000 Or we need to define some default images that will appear if the user doesn't upload an image. 222 00:17:47,000 --> 00:17:49,000 What should be the default image appearing? 223 00:17:50,000 --> 00:17:53,000 I will set some image configurations over here. 224 00:17:58,000 --> 00:18:03,000 So first we need to create a directory by the name images over here. 225 00:18:11,000 --> 00:18:14,000 And inside this directory I will add some free images. 226 00:18:16,000 --> 00:18:21,000 So this is the default detected image that will appear in our Streamlit application. 227 00:18:21,000 --> 00:18:23,000 This is the image one. 228 00:18:23,000 --> 00:18:24,000 This is the default image as well. 229 00:18:24,000 --> 00:18:28,000 And this is another image which will appear in our images directory. 230 00:18:29,000 --> 00:18:29,000 Okay. 231 00:18:34,000 --> 00:18:40,000 So in the root over here we have defined a relative path of the root directory with respect to the current 232 00:18:40,000 --> 00:18:41,000 working Writing directory. 233 00:18:50,000 --> 00:18:54,000 And we need to define our default image over here. 234 00:18:56,000 --> 00:18:58,000 So we can go to images directory. 235 00:18:58,000 --> 00:19:00,000 And over here. 236 00:19:03,000 --> 00:19:05,000 We have image one or jpg. 237 00:19:08,000 --> 00:19:12,000 Then we have default detect image. 238 00:19:20,000 --> 00:19:26,000 And we have detected image one dot okay. 239 00:19:30,000 --> 00:19:33,000 Similarly we can do for the video as well. 240 00:19:33,000 --> 00:19:37,000 For this I will just create another directory where the name videos over here. 241 00:19:47,000 --> 00:19:50,000 And over here I will add two videos. 242 00:19:52,000 --> 00:19:52,000 Okay. 243 00:20:03,000 --> 00:20:10,000 Well first define video directory is equal to root and videos for the name of the directory is videos 244 00:20:11,000 --> 00:20:12,000 okay. 245 00:20:16,000 --> 00:20:18,000 And then we will create a dictionary. 246 00:20:25,000 --> 00:20:26,000 Right. 247 00:20:26,000 --> 00:20:27,000 Video one. 248 00:20:31,000 --> 00:20:33,000 Video directory. 249 00:20:36,000 --> 00:20:37,000 Video one dot. 250 00:20:42,000 --> 00:20:44,000 Then we will write video two. 251 00:20:47,000 --> 00:20:52,000 In your directory media two dot mp4. 252 00:20:53,000 --> 00:20:53,000 Okay. 253 00:20:57,000 --> 00:21:02,000 Next we need to define model configurations where we will create a weight folder over here. 254 00:21:02,000 --> 00:21:08,000 And I will add my uh object detection instance segmentation and pose estimation over 11 models. 255 00:21:08,000 --> 00:21:10,000 So I will just write model configuration. 256 00:21:18,000 --> 00:21:26,000 So first what I will do is I will just create a new directory by the name weights over here. 257 00:21:27,000 --> 00:21:32,000 And inside this I will add what I will do over here is uh we can go over here. 258 00:21:34,000 --> 00:21:34,000 Okay. 259 00:21:34,000 --> 00:21:37,000 So you can just go to your 11 GitHub. 260 00:21:37,000 --> 00:21:37,000 up. 261 00:21:39,000 --> 00:21:46,000 And if you want to download any object detection model or instance segmentation or pose estimation model, 262 00:21:46,000 --> 00:21:51,000 uh, you can simply click over here and you can see that it start downloading. 263 00:21:51,000 --> 00:21:55,000 And if you want to download any segmentation model, you can simply click over here. 264 00:21:55,000 --> 00:21:57,000 And it will now start downloading. 265 00:21:57,000 --> 00:22:01,000 And if you want to download Pose Estimation model you can simply click over here. 266 00:22:01,000 --> 00:22:02,000 It will start downloading okay. 267 00:22:05,000 --> 00:22:08,000 So I have already downloaded so I'm not downloading again. 268 00:22:15,000 --> 00:22:18,000 So I'm just adding these three models over here. 269 00:22:20,000 --> 00:22:25,000 So you can see we have your 11 pose estimation model and your 11 segmentation model. 270 00:22:29,000 --> 00:22:30,000 So I will just write route. 271 00:22:30,000 --> 00:22:34,000 And you just need to go to the weights directory over here. 272 00:22:42,000 --> 00:22:44,000 So we just need to go to model directory. 273 00:22:44,000 --> 00:22:48,000 And here we have the detection model as your 11 nano dash. 274 00:22:48,000 --> 00:22:52,000 So I go to 11 nano that we got it. 275 00:22:57,000 --> 00:23:01,000 So in case of your custom model. 276 00:23:07,000 --> 00:23:08,000 You will write detection. 277 00:23:12,000 --> 00:23:13,000 Dash model. 278 00:23:13,000 --> 00:23:17,000 And here you will write model dash directory. 279 00:23:17,000 --> 00:23:25,000 And over here you will add custom model with dot 80. 280 00:23:25,000 --> 00:23:28,000 And you will just comment this line of code okay. 281 00:23:28,000 --> 00:23:29,000 And uncomment this okay. 282 00:23:32,000 --> 00:23:36,000 Next we need to define the segmentation model. 283 00:23:48,000 --> 00:23:51,000 Next we need to also define the pose estimation model. 284 00:23:59,000 --> 00:24:00,000 Okay. 285 00:24:00,000 --> 00:24:01,000 So that things work. 286 00:24:06,000 --> 00:24:09,000 So we are good for now. 287 00:24:15,000 --> 00:24:18,000 Next uh, if the user uh we need to add the code. 288 00:24:18,000 --> 00:24:22,000 If the user select the detection segmentation or the pose estimation model. 289 00:24:37,000 --> 00:24:39,000 Now if the model die base. 290 00:24:45,000 --> 00:24:47,000 Or a model die based detection. 291 00:24:49,000 --> 00:24:52,000 Model path is equal to path. 292 00:24:55,000 --> 00:24:57,000 Detection model as if. 293 00:25:02,000 --> 00:25:03,000 If the model die based. 294 00:25:03,000 --> 00:25:05,000 Segmentation. 295 00:25:07,000 --> 00:25:09,000 Model path. 296 00:25:10,000 --> 00:25:22,000 Path segmentation model and if model type is equal to for estimation. 297 00:25:27,000 --> 00:25:30,000 On the label or. 298 00:25:32,000 --> 00:25:33,000 For estimation. 299 00:25:33,000 --> 00:25:33,000 On. 300 00:25:42,000 --> 00:25:44,000 So next we will load the YOLO model. 301 00:25:46,000 --> 00:25:48,000 When I try. 302 00:25:50,000 --> 00:25:53,000 YOLO more than five. 303 00:25:56,000 --> 00:25:59,000 And write except exception as e. 304 00:26:02,000 --> 00:26:04,000 S g dot error. 305 00:26:07,000 --> 00:26:12,000 You can see the write unable to load model. 306 00:26:12,000 --> 00:26:15,000 Check uh s file path. 307 00:26:21,000 --> 00:26:21,000 Okay. 308 00:26:27,000 --> 00:26:29,000 Simply write this over here. 309 00:26:32,000 --> 00:26:36,000 And we can also display the error in our Streamlit application. 310 00:26:42,000 --> 00:26:46,000 Now we can just set image and video configuration. 311 00:27:05,000 --> 00:27:10,000 Will give the user the option to select if you want to perform object detection, instance segmentation 312 00:27:10,000 --> 00:27:13,000 and pose estimation on image or on video. 313 00:27:13,000 --> 00:27:15,000 So we'll give the user the option. 314 00:27:23,000 --> 00:27:26,000 We can start sidebar dot radio. 315 00:27:29,000 --> 00:27:30,000 Like source. 316 00:27:34,000 --> 00:27:35,000 Source list. 317 00:27:39,000 --> 00:27:40,000 Sorry. 318 00:27:40,000 --> 00:27:43,000 Initialize source image is equal to none. 319 00:27:46,000 --> 00:27:47,000 So if that. 320 00:27:47,000 --> 00:27:51,000 If the user selects an image. 321 00:27:59,000 --> 00:28:01,000 We will ask the user to upload an image. 322 00:28:16,000 --> 00:28:20,000 And the Streamlit application will accept the image of type. 323 00:28:24,000 --> 00:28:25,000 JPG png. 324 00:28:28,000 --> 00:28:29,000 JPEG. 325 00:28:31,000 --> 00:28:32,000 BMP. 326 00:28:36,000 --> 00:28:38,000 So I only know all these image types. 327 00:28:38,000 --> 00:28:42,000 If you know some other, you can also add it over here as well. 328 00:28:44,000 --> 00:28:48,000 Next we will create two columns which will contain a default image and other will contain the detected 329 00:28:48,000 --> 00:28:52,000 image, or one will contain the. 330 00:28:52,000 --> 00:28:57,000 Use the image that the user has uploaded and other is the our output image with the detections. 331 00:29:03,000 --> 00:29:05,000 So first we will go towards the column one. 332 00:29:12,000 --> 00:29:17,000 So if the source image is none like the user doesn't upload an image. 333 00:29:23,000 --> 00:29:26,000 Then we will define a default image path over here. 334 00:29:37,000 --> 00:29:39,000 And we will say. 335 00:29:42,000 --> 00:29:43,000 This is our default image. 336 00:29:57,000 --> 00:29:58,000 You can also open this image. 337 00:30:10,000 --> 00:30:14,000 And we can also display this image in our Streamlit application using S3 dot image. 338 00:30:16,000 --> 00:30:17,000 Right. 339 00:30:17,000 --> 00:30:18,000 Default image path. 340 00:30:21,000 --> 00:30:24,000 And our caption will be before image. 341 00:30:48,000 --> 00:30:48,000 Else. 342 00:30:48,000 --> 00:30:55,000 If the user uploads an image and source image is not on, then we will display the uploaded image. 343 00:30:59,000 --> 00:31:06,000 So if the user uploads an image, then we will display the source image. 344 00:31:09,000 --> 00:31:13,000 Caption will be uploaded. 345 00:31:13,000 --> 00:31:14,000 Image. 346 00:31:22,000 --> 00:31:22,000 Okay. 347 00:31:25,000 --> 00:31:27,000 And now we will add accept as well. 348 00:31:27,000 --> 00:31:32,000 Except exception as okay. 349 00:31:36,000 --> 00:31:37,000 Okay. 350 00:31:38,000 --> 00:31:39,000 What will be. 351 00:31:42,000 --> 00:31:43,000 Any. 352 00:31:53,000 --> 00:31:54,000 Error occurred. 353 00:31:56,000 --> 00:31:57,000 While. 354 00:32:03,000 --> 00:32:06,000 Opening that and we will. 355 00:32:08,000 --> 00:32:10,000 Display the error as. 356 00:32:13,000 --> 00:32:16,000 Now we'll create another column as well. 357 00:32:31,000 --> 00:32:33,000 But if forcing it is not. 358 00:32:36,000 --> 00:32:42,000 Then we'll define default detected image path over here. 359 00:33:03,000 --> 00:33:04,000 Okay. 360 00:33:10,000 --> 00:33:14,000 And here we'll pass the default detected image. 361 00:33:16,000 --> 00:33:17,000 Path. 362 00:33:19,000 --> 00:33:21,000 Caption will be detected. 363 00:33:21,000 --> 00:33:22,000 Image. 364 00:33:39,000 --> 00:33:49,000 Else if the source image is there, and user also has a sidebar button which is. 365 00:33:51,000 --> 00:33:53,000 Detect objects. 366 00:33:57,000 --> 00:34:02,000 That we will use model dot predict. 367 00:34:03,000 --> 00:34:04,000 Okay. 368 00:34:05,000 --> 00:34:08,000 And here we will pass the uploaded image. 369 00:34:11,000 --> 00:34:12,000 And. 370 00:34:16,000 --> 00:34:18,000 The first uploaded image over here. 371 00:34:19,000 --> 00:34:20,000 Then. 372 00:34:28,000 --> 00:34:32,000 But we also define the confidence value. 373 00:34:36,000 --> 00:34:37,000 It'sand. 374 00:34:42,000 --> 00:34:45,000 That's not the result as well. 375 00:35:25,000 --> 00:35:30,000 And we'll also show the bounding box coordinates there as well. 376 00:35:31,000 --> 00:35:32,000 Right. 377 00:35:32,000 --> 00:35:34,000 With a dot expander. 378 00:36:02,000 --> 00:36:04,000 And we will display the error. 379 00:36:08,000 --> 00:36:13,000 And we can also add section over here as well. 380 00:36:27,000 --> 00:36:28,000 Worthwhile. 381 00:36:35,000 --> 00:36:35,000 Okay. 382 00:36:35,000 --> 00:36:36,000 That works. 383 00:36:38,000 --> 00:36:43,000 Now, let's test, uh, all our Streamlit application works on image. 384 00:36:43,000 --> 00:36:48,000 And if it works fine, then we'll go ahead with the video as well. 385 00:36:52,000 --> 00:36:55,000 Okay, so let's test our Streamlit application. 386 00:36:58,000 --> 00:37:03,000 So here you can see that this is a Streamlit application that we have created so far. 387 00:37:03,000 --> 00:37:08,000 So first uh, let's reduction I will just upload this image. 388 00:37:10,000 --> 00:37:12,000 And if I click on detect object from here. 389 00:37:12,000 --> 00:37:14,000 So this is the uploaded image. 390 00:37:14,000 --> 00:37:15,000 Okay. 391 00:37:15,000 --> 00:37:19,000 So now you can see over here we are able to detect the person purse backpack. 392 00:37:19,000 --> 00:37:24,000 And here you can see we have the bounding box coordinates for each detected object in the form of a 393 00:37:24,000 --> 00:37:25,000 tensor. 394 00:37:25,000 --> 00:37:29,000 We can also do segmentation and let's click on detect object from here. 395 00:37:31,000 --> 00:37:34,000 So now you can see that we are able to form instance segmentation. 396 00:37:34,000 --> 00:37:38,000 And we have drawn mask around each of the detected object. 397 00:37:38,000 --> 00:37:40,000 We can also perform pose estimation. 398 00:37:40,000 --> 00:37:43,000 And let's click on detect object from here. 399 00:37:46,000 --> 00:37:49,000 So now you can see we are able to perform pose estimation. 400 00:37:49,000 --> 00:37:52,000 And we have drawn key points around each of the detected object. 401 00:37:52,000 --> 00:37:58,000 And here you can see we have the bounding box coordinates appearing over here for each of the object. 402 00:37:58,000 --> 00:38:00,000 Okay so we are done with the image part. 403 00:38:00,000 --> 00:38:05,000 Let's go ahead and move towards the uh video part as well. 404 00:38:05,000 --> 00:38:06,000 Okay. 405 00:38:12,000 --> 00:38:13,000 So we can write and if. 406 00:38:16,000 --> 00:38:20,000 Source radio is equal to radio. 407 00:38:22,000 --> 00:38:23,000 Copy. 408 00:38:23,000 --> 00:38:23,000 Right. 409 00:38:23,000 --> 00:38:32,000 Source video is equal to ask not sidebar or select box. 410 00:38:38,000 --> 00:38:40,000 It was a video. 411 00:38:50,000 --> 00:38:52,000 In your dictionary dot keys. 412 00:38:59,000 --> 00:39:02,000 And right will open with your dictionary. 413 00:39:05,000 --> 00:39:10,000 And what source video or read it. 414 00:39:13,000 --> 00:39:24,000 As I so we need to convert the video into bytes so we can perform object detection on the video. 415 00:39:37,000 --> 00:39:41,000 Or S3 dot video will display the video into our Streamlit application. 416 00:39:56,000 --> 00:39:59,000 So if the user click on this button right here. 417 00:40:02,000 --> 00:40:02,000 And. 418 00:40:09,000 --> 00:40:12,000 Then we'll add a try and accept permission. 419 00:40:34,000 --> 00:40:35,000 Okay. 420 00:40:42,000 --> 00:40:44,000 So we'll create an empty frame from now. 421 00:40:48,000 --> 00:40:51,000 Now we'll just create a while loop over here. 422 00:41:11,000 --> 00:41:14,000 And if we are able to read the frame then. 423 00:41:16,000 --> 00:41:19,000 We'll just resize the frame over here. 424 00:41:48,000 --> 00:41:55,000 Next we will predict the objects in the image using YOLO 11. 425 00:41:57,000 --> 00:41:59,000 Right result is equal to model dot. 426 00:41:59,000 --> 00:42:06,000 Predict image is equal to confidence is equal to confidence value. 427 00:42:10,000 --> 00:42:16,000 Next we will plot the detected objects on the radio Rain. 428 00:42:34,000 --> 00:42:36,000 Uh, we already have created an empty frame. 429 00:42:36,000 --> 00:42:39,000 We will be plotting our detected objects on that frame. 430 00:43:16,000 --> 00:43:23,000 And if the video frames end, then we'll release all the frames and we will break the while loop. 431 00:43:25,000 --> 00:43:29,000 And we can also add an accept conditional here. 432 00:43:46,000 --> 00:43:48,000 Around loading video. 433 00:43:53,000 --> 00:43:58,000 Okay so and we will display the error over here as well. 434 00:43:59,000 --> 00:44:02,000 So now we have created our Streamlit application. 435 00:44:02,000 --> 00:44:06,000 Now it's time to test our Streamlit app application. 436 00:44:06,000 --> 00:44:11,000 So we can simply write over here Streamlit run main.py. 437 00:44:12,000 --> 00:44:20,000 Okay so let's test our Streamlit application and let's see if we are able to do detection on image and 438 00:44:20,000 --> 00:44:21,000 video as well. 439 00:44:21,000 --> 00:44:23,000 We have tested for image already now. 440 00:44:23,000 --> 00:44:26,000 Now we will mainly test on video now. 441 00:44:31,000 --> 00:44:34,000 So first of all I will just click on video from here. 442 00:44:34,000 --> 00:44:41,000 And um I can just call it video two and click on Detect Video Objects from here. 443 00:44:41,000 --> 00:44:46,000 And now you can see over here we are able to detect a person bicycle from here. 444 00:44:46,000 --> 00:44:49,000 And the detection results look quite promising. 445 00:44:50,000 --> 00:44:53,000 If you want to perform segmentation you can select segmentation. 446 00:44:53,000 --> 00:44:53,000 From here. 447 00:44:53,000 --> 00:44:57,000 You can select any video that you have saved in the video directory. 448 00:44:57,000 --> 00:45:00,000 And then you will run Detect video objects. 449 00:45:00,000 --> 00:45:03,000 Now you can see that we are able to perform instance segmentation. 450 00:45:03,000 --> 00:45:06,000 We are able to detect a traffic lights bicycle person. 451 00:45:06,000 --> 00:45:11,000 And you can see over here we have drawn a mask around each of the detected object. 452 00:45:11,000 --> 00:45:12,000 So this is what we do. 453 00:45:12,000 --> 00:45:13,000 In instance segmentation. 454 00:45:13,000 --> 00:45:18,000 In instance segmentation, we draw a mask around each of the detected object. 455 00:45:18,000 --> 00:45:24,000 Similarly, we can perform pose estimation and they can detect 3D objects from here. 456 00:45:24,000 --> 00:45:26,000 So now you can see we have performed pose estimation. 457 00:45:26,000 --> 00:45:30,000 And we have drawn key points around each of the detected object. 458 00:45:30,000 --> 00:45:35,000 Similarly you can adjust the confidence threshold to 79 as well. 459 00:45:35,000 --> 00:45:40,000 And you can see now as I increase the confidence threshold the detection has decreased. 460 00:45:40,000 --> 00:45:42,000 But now the detections are more accurate. 461 00:45:42,000 --> 00:45:47,000 But if I decrease the confidence score there can be some false positives as well. 462 00:45:47,000 --> 00:45:49,000 But there have been more number of detections. 463 00:45:51,000 --> 00:45:53,000 Similarly you can do it for image as well. 464 00:45:53,000 --> 00:45:55,000 We have already tried on images. 465 00:45:58,000 --> 00:45:59,000 And detect objects. 466 00:46:01,000 --> 00:46:05,000 And let me see if I have some other image over here as well. 467 00:46:08,000 --> 00:46:12,000 Let's use this image and let's click on detect objects. 468 00:46:12,000 --> 00:46:15,000 So now you can see over here we are able to detect the objects. 469 00:46:15,000 --> 00:46:20,000 And we are also able to draw the key points around each of the detected object because we have selected 470 00:46:20,000 --> 00:46:21,000 for the estimation. 471 00:46:21,000 --> 00:46:28,000 And if I just like segmentation, you can see we are able to draw a mask around each of the detected. 472 00:46:31,000 --> 00:46:37,000 So now you can see over here with the help of instance segmentation, we are able to draw a mask around 473 00:46:37,000 --> 00:46:38,000 each of detected object. 474 00:46:38,000 --> 00:46:43,000 And if I click on detection from here and click on detect object. 475 00:46:43,000 --> 00:46:47,000 From here you can see that I am able to detect a different objects like person bus. 476 00:46:47,000 --> 00:46:52,000 And over here if I just expand this I have all the bounding box coordinate informations in the form 477 00:46:52,000 --> 00:46:54,000 of tensors over here. 478 00:46:54,000 --> 00:46:56,000 So that's all from this tutorial. 479 00:46:56,000 --> 00:47:01,000 In this tutorial we have learned how we can create a Streamlit application to do object detection, 480 00:47:01,000 --> 00:47:07,000 instance segmentation and pose estimation with the help of ultralytics YOLO 11 on images as well as 481 00:47:07,000 --> 00:47:08,000 on videos. 482 00:47:08,000 --> 00:47:09,000 That's all from this video. 483 00:47:09,000 --> 00:47:11,000 Thank you for watching.