1 00:00:02,580 --> 00:00:05,670 But now we will see how we can integrate YOLO v9 with flask. 2 00:00:05,670 --> 00:00:14,610 So I will just go over here and create a new uh.py file by the name last app.py. 3 00:00:16,690 --> 00:00:21,070 Okay, so here I will just writing all my code to create a flask app. 4 00:00:22,090 --> 00:00:28,660 So first of all we will import all the the required libraries over here. 5 00:00:33,690 --> 00:00:37,560 So I am just importing all the required libraries. 6 00:00:40,320 --> 00:00:47,490 I would just like, uh, before this, we need to install the last package, so just install last. 7 00:00:47,940 --> 00:00:48,480 Okay. 8 00:00:51,540 --> 00:00:53,970 Here I will just write from flask import. 9 00:00:55,670 --> 00:00:59,000 Thus rendered template. 10 00:00:59,000 --> 00:01:00,770 We will be using this library later on. 11 00:01:03,110 --> 00:01:04,400 Response. 12 00:01:12,900 --> 00:01:15,900 There Sonify requests. 13 00:01:18,730 --> 00:01:20,170 Ah, we don't need to disturb. 14 00:01:20,470 --> 00:01:20,920 Okay. 15 00:01:22,960 --> 00:01:26,020 Then we can just import OS and. 16 00:01:27,200 --> 00:01:30,380 Import C window opens with Python package. 17 00:01:30,380 --> 00:01:34,340 So in the meanwhile our class library has been installed, so that's good. 18 00:01:36,770 --> 00:01:39,500 Last from this detection.py file. 19 00:01:39,500 --> 00:01:43,400 We will be importing this object detection function as well. 20 00:01:43,820 --> 00:01:52,460 So I will just write over here from detection import object detection okay so that's good. 21 00:01:53,900 --> 00:01:59,720 So in the step number one we need to initialize the class application. 22 00:02:01,250 --> 00:02:05,300 So to initialize the flask application I will just write over here. 23 00:02:06,470 --> 00:02:09,350 Now you can see that we have initialized the flask application. 24 00:02:10,220 --> 00:02:10,580 Okay. 25 00:02:10,580 --> 00:02:12,980 So now next we will configure. 26 00:02:14,510 --> 00:02:16,340 Biosecurity. 27 00:02:19,370 --> 00:02:22,100 Well, we'll just write one app dot config. 28 00:02:28,870 --> 00:02:29,230 Bigger. 29 00:02:29,500 --> 00:02:31,900 And you can write any key over here. 30 00:02:31,900 --> 00:02:33,310 That's not necessary. 31 00:02:34,390 --> 00:02:37,840 So I will just write my own name over here. 32 00:02:43,310 --> 00:02:46,430 So as you know that we are creating a web app using class. 33 00:02:46,430 --> 00:02:54,260 So we will create an upload folder, uh, where if the user uploads any input video image so that uh 34 00:02:54,260 --> 00:02:56,390 file will be saved into that folder. 35 00:02:59,250 --> 00:03:01,950 Boredstalker uploaded. 36 00:03:15,670 --> 00:03:18,460 So for this I will just write app dot one. 37 00:03:18,460 --> 00:03:18,970 Sick. 38 00:03:23,170 --> 00:03:24,520 Upload folder. 39 00:03:29,130 --> 00:03:29,670 That date. 40 00:03:31,960 --> 00:03:35,740 Well, but you can see that we have not created this folder anywhere over here. 41 00:03:35,740 --> 00:03:38,950 So I will just create a new directory that date. 42 00:03:40,090 --> 00:03:40,450 Okay. 43 00:03:40,450 --> 00:03:46,120 And over here I will just create another decade cycles okay. 44 00:03:46,120 --> 00:03:48,670 So now you can see that we have created the required directories. 45 00:03:52,990 --> 00:04:01,690 So now we will just create another function by the name generate frames, and the input to this function 46 00:04:01,690 --> 00:04:08,620 will be the input file path, whether it can be image image audio okay. 47 00:04:08,620 --> 00:04:09,130 So. 48 00:04:12,270 --> 00:04:14,430 Or it can be for live webcam feed. 49 00:04:14,430 --> 00:04:16,020 You can pass it over here. 50 00:04:17,760 --> 00:04:18,390 So. 51 00:04:21,810 --> 00:04:23,910 But within this right, you'll win an output. 52 00:04:23,910 --> 00:04:27,810 And from this object detection function which we have imported. 53 00:04:27,810 --> 00:04:31,560 So let's write object detection action. 54 00:04:33,090 --> 00:04:35,490 So now you can see this object detection function. 55 00:04:35,490 --> 00:04:41,490 Take only the input file as an input whether it can be image video. 56 00:04:42,210 --> 00:04:47,460 So the input file path will be passed uh to this function as input okay. 57 00:04:47,460 --> 00:04:49,140 So input file path. 58 00:04:57,410 --> 00:05:06,500 So and this function of basically yields I to image uh or current frame or then the frame rate in which 59 00:05:06,500 --> 00:05:09,080 the shape of the current frame and total number of iterations. 60 00:05:09,080 --> 00:05:14,480 So these are the four outputs I am zero frame. 61 00:05:47,640 --> 00:05:51,510 So now you can see that we have the encoded image over here. 62 00:05:53,630 --> 00:05:55,700 Is not there should be in code. 63 00:05:55,730 --> 00:05:55,940 Okay. 64 00:05:55,940 --> 00:05:59,180 So now you can see that, uh, we have the encoded image over here. 65 00:05:59,180 --> 00:06:05,210 So any flask application basically requires the encoded image, uh, to be converted into bytes and 66 00:06:05,210 --> 00:06:07,070 rendered into the HTML. 67 00:06:07,070 --> 00:06:08,030 So. 68 00:06:09,620 --> 00:06:11,960 To convert this encoded image to into bytes. 69 00:06:11,960 --> 00:06:15,230 We'll be using dot two bytes uh function. 70 00:06:15,260 --> 00:06:21,950 So what I will do over here is I will write frame is equal to buffer dot two bytes. 71 00:06:23,770 --> 00:06:25,060 So two bytes. 72 00:06:25,090 --> 00:06:25,720 Dot two bytes. 73 00:06:25,720 --> 00:06:27,880 Convert the encoded image into bytes. 74 00:06:29,940 --> 00:06:30,390 That. 75 00:06:32,700 --> 00:06:38,610 And using dot gain keyword like you can see over here we will display the individual frames. 76 00:06:38,760 --> 00:06:39,420 Okay. 77 00:06:50,880 --> 00:06:58,530 And, uh, we will loop through all the individual frames and, uh, go through all the individual frames 78 00:06:58,740 --> 00:07:05,430 and to convert the individual frames are into subsequent frames will use the content type. 79 00:07:05,670 --> 00:07:10,050 So we need to convert all these individual frames into subsequent frames okay. 80 00:07:13,800 --> 00:07:14,580 It's all. 81 00:07:21,010 --> 00:07:26,560 Right content type so that I can convert the individual frames into subsequent frames. 82 00:07:57,170 --> 00:07:58,820 Okay, so that's good. 83 00:07:58,820 --> 00:07:59,930 We have done good enough. 84 00:07:59,930 --> 00:08:02,150 So we have created a generate frame function. 85 00:08:02,150 --> 00:08:07,100 I'm not currently calculating the frame rate frame shape because I don't require this yet. 86 00:08:11,590 --> 00:08:15,730 Then I will just write app dot route so that I can create a URL. 87 00:08:23,210 --> 00:08:24,440 That you can just. 88 00:08:27,290 --> 00:08:28,370 Let's try it out. 89 00:08:28,370 --> 00:08:29,390 This as well. 90 00:08:44,500 --> 00:08:45,160 That films. 91 00:08:45,160 --> 00:08:48,730 And here we will be calling the generate frame function, which are created above. 92 00:09:03,950 --> 00:09:06,530 So currently I will just pause the video part. 93 00:09:06,530 --> 00:09:09,380 So I will just running video two dot mp4. 94 00:09:42,510 --> 00:09:44,040 Okay, so that's all. 95 00:09:44,040 --> 00:09:45,840 I don't see any error yet. 96 00:09:45,840 --> 00:09:48,330 So just right over here. 97 00:09:59,930 --> 00:10:03,770 And out run so that we can make some changes in the runtime. 98 00:10:03,770 --> 00:10:08,690 So I've just set debug is equal to true so that we can make changes in the runtime. 99 00:10:09,500 --> 00:10:13,790 So let's get started with this. 100 00:10:17,550 --> 00:10:21,120 So I will just open the terminal now and I will just write over here. 101 00:10:21,120 --> 00:10:21,900 Python. 102 00:10:22,800 --> 00:10:27,450 Ask app dot Pi, and let's see if there is an error so that we can fix this up. 103 00:10:38,750 --> 00:10:43,760 You can similarly pass an input image path over here or zero for the live webcam feed. 104 00:10:43,880 --> 00:10:45,080 Let's open this URL. 105 00:10:55,770 --> 00:11:00,240 But this will take some time before we have the results over here. 106 00:11:06,040 --> 00:11:09,220 The legislature bought seconds. 107 00:11:09,220 --> 00:11:14,410 I'm not going to pause the video so we can see how many time lag we get. 108 00:11:37,280 --> 00:11:39,710 So this will take a few more seconds. 109 00:11:39,710 --> 00:11:40,190 Get. 110 00:11:59,400 --> 00:12:07,320 Okay, so the processing on the frames have started, but why not just appearing over here? 111 00:12:19,910 --> 00:12:22,730 I can see over here the perception. 112 00:12:26,870 --> 00:12:28,520 I will be starting this. 113 00:12:42,090 --> 00:12:42,540 Oh, yeah. 114 00:12:42,540 --> 00:12:43,410 It's working now. 115 00:12:43,410 --> 00:12:43,800 Yeah. 116 00:12:44,310 --> 00:12:45,030 So that's great. 117 00:12:45,030 --> 00:12:49,260 So now you can see over here, uh, we have the URL over here. 118 00:12:49,260 --> 00:12:52,890 I've just passed this URL and this is the frame size which we are getting. 119 00:12:53,100 --> 00:13:00,150 Um, the width of this frame is 12, uh, 80 and the height is uh, 720, which you can find over here. 120 00:13:00,150 --> 00:13:03,720 And you can see that we are able to track different cars over here. 121 00:13:03,870 --> 00:13:07,860 Uh, I'm testing on CPU, so the processing is very, very slow. 122 00:13:07,860 --> 00:13:13,620 But, uh, if you have a good GPU and a processing will be very fast, like, you can see the car is 123 00:13:13,620 --> 00:13:18,990 moving, but the as the processing or frame rate is very low over here. 124 00:13:18,990 --> 00:13:22,980 So now you can see that we are able to to churn on this video. 125 00:13:22,980 --> 00:13:24,990 We have written different cars over here. 126 00:13:24,990 --> 00:13:25,890 This is a wrong detection. 127 00:13:25,890 --> 00:13:26,670 This is not a truck. 128 00:13:26,670 --> 00:13:28,950 And these are also false positive. 129 00:13:28,950 --> 00:13:31,350 Like these are the wrong detections okay. 130 00:13:31,350 --> 00:13:39,150 So in the same way uh you can pass uh like you don't need to close up this program simply, uh, we 131 00:13:39,150 --> 00:13:43,230 can also do object detection over an image like click or. 132 00:13:43,230 --> 00:13:43,440 Okay. 133 00:13:43,440 --> 00:13:44,400 So let's see. 134 00:13:47,130 --> 00:13:54,180 You just need to refresh this up and let's see if we are able to detect different objects in this image 135 00:13:54,180 --> 00:13:54,900 or not. 136 00:14:02,680 --> 00:14:07,360 It is, though, it's worth a few more seconds. 137 00:14:08,200 --> 00:14:11,110 Okay, so now here you can see we have our output image. 138 00:14:11,110 --> 00:14:14,230 We are able to detect this talk person this handbag. 139 00:14:14,440 --> 00:14:15,070 Okay. 140 00:14:15,070 --> 00:14:18,130 And let's test out on other image. 141 00:14:20,620 --> 00:14:21,370 Image. 142 00:14:22,420 --> 00:14:25,210 So let's try this out. 143 00:14:25,990 --> 00:14:26,140 Okay. 144 00:14:26,140 --> 00:14:28,270 So I will just click on refresh from here. 145 00:14:28,750 --> 00:14:32,770 And let's see if we are able to detect different objects in this image or not. 146 00:14:32,770 --> 00:14:40,540 Like you can see over here we are able to score a person dog and bag. 147 00:14:40,630 --> 00:14:40,900 So. 148 00:14:43,850 --> 00:14:50,150 The results are very good overall, but let's see if it works on this other image or not. 149 00:14:51,440 --> 00:14:55,370 So as I'm trying it on, CPU processing is very low. 150 00:14:55,370 --> 00:14:57,290 But you can see here we have the results. 151 00:14:57,290 --> 00:15:04,280 We are able to detect this bus, traffic lights, person bag, bag, bicycle, hand bag, all this stuff 152 00:15:04,280 --> 00:15:07,340 are being detected and you can see this person are very far away. 153 00:15:07,340 --> 00:15:08,150 Plus they are blurred. 154 00:15:08,150 --> 00:15:10,670 But we are also able to track these persons as well. 155 00:15:10,670 --> 00:15:12,890 And we have both the traffic lights as well. 156 00:15:13,190 --> 00:15:16,580 So you can try it out on some other video as well. 157 00:15:16,670 --> 00:15:21,560 Like uh video tool dot uh mp4. 158 00:15:22,160 --> 00:15:23,090 Let's try it out. 159 00:15:23,090 --> 00:15:23,420 This. 160 00:15:28,100 --> 00:15:31,100 Similarly, you can also test it on the live webcam feed as well. 161 00:15:31,130 --> 00:15:32,750 I'll show you as well. 162 00:15:34,790 --> 00:15:37,580 Are you going to try it out on the live webcam feed? 163 00:15:44,600 --> 00:15:50,600 So it's Wednesday full time and you can see here we have the frame size for this video. 164 00:15:50,630 --> 00:15:51,170 Okay. 165 00:15:52,310 --> 00:15:55,490 So let's see what is it. 166 00:15:55,550 --> 00:15:59,750 So now you can see that we are able to detect different persons in this video. 167 00:16:00,140 --> 00:16:00,650 Okay. 168 00:16:00,650 --> 00:16:03,170 Like you can see in the processing is very slow. 169 00:16:03,200 --> 00:16:06,380 I'm using CPU else everything is okay. 170 00:16:06,410 --> 00:16:10,100 So in this way you can see that we are doing object detection on the video. 171 00:16:10,100 --> 00:16:14,030 And if you want to do it on the live webcam feed, just change it to zero. 172 00:16:17,360 --> 00:16:18,980 And let's run this up now. 173 00:16:31,380 --> 00:16:33,060 So this will take some time. 174 00:16:44,230 --> 00:16:44,530 Okay. 175 00:16:45,580 --> 00:16:46,660 Okay. 176 00:16:46,960 --> 00:16:47,920 Let's see. 177 00:16:47,920 --> 00:16:49,030 How does it goes? 178 00:16:57,490 --> 00:17:01,270 Uh, this will take a few more seconds, uh, before we have the results. 179 00:17:14,280 --> 00:17:18,870 Okay, so now you can see our fear of the live webcam feed is on. 180 00:17:19,590 --> 00:17:20,160 Okay. 181 00:17:20,160 --> 00:17:22,830 And you can find me young. 182 00:17:25,440 --> 00:17:31,110 But okay, so now you can see over here, uh, we are able to track the person as well. 183 00:17:31,140 --> 00:17:31,830 Over here. 184 00:17:31,830 --> 00:17:33,210 This is a false positive. 185 00:17:33,210 --> 00:17:35,160 Like you can see this is not the chair. 186 00:17:35,250 --> 00:17:40,860 So in this way you can do object detection with the, like webcam feed over here as well. 187 00:17:42,050 --> 00:17:42,470 Okay. 188 00:17:42,470 --> 00:17:43,070 So. 189 00:17:46,710 --> 00:17:48,630 So this is how it goes. 190 00:17:48,960 --> 00:17:55,230 Okay, so in this tutorial we have seen that how we can integrate YOLO v9 with flask and how we can 191 00:17:55,230 --> 00:18:00,930 do object detection on image video and on the live webcam feed. 192 00:18:00,930 --> 00:18:02,910 And that's all for this tutorial. 193 00:18:02,910 --> 00:18:03,870 Thank you for watching.