Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:00,510 --> 00:00:05,880 Hello and welcome to listen to toil to they we're going to do the face recognition with the web cam 2 00:00:06,210 --> 00:00:13,350 by playing the detect function that we made in the previous toils on each of the images coming from 3 00:00:13,350 --> 00:00:14,520 the webcam. 4 00:00:14,520 --> 00:00:14,850 All right. 5 00:00:14,850 --> 00:00:16,050 So let's do this. 6 00:00:16,050 --> 00:00:21,680 And if that doesn't take a while we might actually see the results at the end of this tutorial. 7 00:00:21,720 --> 00:00:27,510 So the first thing we need is the last frame coming from the webcam because that's on this last frame 8 00:00:27,510 --> 00:00:33,930 that we will apply the detect function and so to get this last frame we're going to create an object 9 00:00:34,080 --> 00:00:36,670 of another class which is video capture. 10 00:00:36,840 --> 00:00:43,260 So that's a class working on the web cam and basically this object will contain several entities several 11 00:00:43,260 --> 00:00:46,260 informations and one of them will be the frame. 12 00:00:46,260 --> 00:00:49,650 The last image last frame coming from the webcam. 13 00:00:49,830 --> 00:00:57,420 So this object is going to be called Video underscore capture because this is actually an object from 14 00:00:58,200 --> 00:01:07,190 the video capture class which is a class of the CB2 much all. 15 00:01:07,200 --> 00:01:14,190 So let's not forget CB2 here video capture class and that class takes one argument actually which is 16 00:01:14,340 --> 00:01:16,220 0 0 1 0. 17 00:01:16,260 --> 00:01:22,140 If it's the webcam of the computer and one if it's a webcam coming from an external device that is if 18 00:01:22,140 --> 00:01:26,040 you have an external webcam that you plugged into your computer. 19 00:01:26,340 --> 00:01:32,460 I'm going to use the webcam of my computer the internal webcam so that's going to be zero. 20 00:01:32,520 --> 00:01:36,330 But if you have an external webcam you will have to put one here. 21 00:01:36,330 --> 00:01:37,260 All right. 22 00:01:37,260 --> 00:01:44,400 So that's basically all we need to do to get this video capture object that contains the last frame 23 00:01:44,580 --> 00:01:45,930 coming from the webcam. 24 00:01:46,170 --> 00:01:54,540 And now we are ready to start the loop to apply to detect function on all the images coming from the 25 00:01:54,540 --> 00:01:59,640 web cam that we get things through this video capture object and on all these images. 26 00:01:59,640 --> 00:02:04,060 We will apply to detect function to detect the face and the eyes. 27 00:02:04,170 --> 00:02:09,270 So that's why we're going to start a loop now which is not going to be a for loop because it's going 28 00:02:09,270 --> 00:02:15,180 to be an infinite loop and we will break it by using the break trick. 29 00:02:15,180 --> 00:02:20,720 And by then you know you can end the loop by just typing break and actually break will be applied. 30 00:02:20,910 --> 00:02:24,920 If we type on q on the keyboard to turn off the webcam. 31 00:02:25,320 --> 00:02:27,590 So this while loop. 32 00:02:27,660 --> 00:02:36,140 The trick is to do well here and then trick which means we repeat infinitely until daybreak. 33 00:02:36,210 --> 00:02:36,720 All right. 34 00:02:36,840 --> 00:02:39,050 So what is the first step in it to do. 35 00:02:39,210 --> 00:02:43,560 Well the first step is to get the last frame coming from the webcam and to get it. 36 00:02:43,620 --> 00:02:50,100 We're going to use a method for more video capture object which is the read method and this read method 37 00:02:50,160 --> 00:02:55,530 returns two elements and the second element is the last frame coming from the webcam. 38 00:02:55,590 --> 00:03:02,700 And since we're only interested in that we're going to use this trick here underscore cummer frame because 39 00:03:02,700 --> 00:03:06,600 the read method returns two elements and we are only interested in this one. 40 00:03:06,600 --> 00:03:11,970 And the fact that we're putting an underscore here means that we won't get the first element returned 41 00:03:11,970 --> 00:03:13,190 by this read method. 42 00:03:14,420 --> 00:03:19,130 All right and so now we need to call this read method to get the last frame of the webcam and to do 43 00:03:19,130 --> 00:03:23,510 this since this method is a method of the video capture class. 44 00:03:23,510 --> 00:03:32,050 Well we need to take our video capture object then that and then read with some parenthesis. 45 00:03:32,060 --> 00:03:32,810 Perfect. 46 00:03:32,990 --> 00:03:39,160 That gets us the last frame of the webcam and now we are almost ready to apply to detect method. 47 00:03:39,170 --> 00:03:44,840 Remember the Digic method not only takes the frame as input but also the gray that is the black and 48 00:03:44,840 --> 00:03:50,210 white version of the frame and therefore we need to do something here to get this black and white version 49 00:03:50,210 --> 00:03:55,100 of this frame that we just got thanks to the read method of our video capture object. 50 00:03:55,280 --> 00:04:03,140 And to do this we're going to use the city color function CVG color is a function of our of each module 51 00:04:03,320 --> 00:04:06,600 that allows to do some transformations on some frames. 52 00:04:06,650 --> 00:04:12,650 So there will be some transformations on our original colored frame and this transformation will be 53 00:04:12,650 --> 00:04:19,190 of course to convert this color image into black and white so that we can get the second argument here 54 00:04:19,610 --> 00:04:22,850 Grey which will be the black and white version of this frame. 55 00:04:23,180 --> 00:04:28,390 So let's call it actually gray black and white version of the frame. 56 00:04:28,530 --> 00:04:36,530 And now we're going to take our city to Mudgal because we want to take the CVT color class which takes 57 00:04:36,920 --> 00:04:42,530 two arguments the first one is of course frame because that's the image we want to convert into black 58 00:04:42,530 --> 00:04:43,260 and white. 59 00:04:43,550 --> 00:04:54,060 And the second one will be civil to that all in capital letters color underscore b g are to gray. 60 00:04:54,560 --> 00:04:56,170 So what is this. 61 00:04:56,180 --> 00:05:01,320 This argument Sivi to that color B.G. are two gray cells. 62 00:05:01,340 --> 00:05:08,060 In fact to do an average on blue green and red and that's to take a scale of gray because you know cause 63 00:05:08,120 --> 00:05:11,310 the Cascades are applied on the gray images. 64 00:05:11,480 --> 00:05:16,670 But in order to get the right contrast of black and white well we need to do some average on the blue 65 00:05:16,760 --> 00:05:19,090 red and green to get this contrast. 66 00:05:19,160 --> 00:05:24,590 You know if we have some parts of the images that are with dark colors and some dark red blue green 67 00:05:24,830 --> 00:05:28,000 Well we will get some dark gray in the great image. 68 00:05:28,010 --> 00:05:32,350 So that's just a trick to obtain the right shades of gray. 69 00:05:32,360 --> 00:05:40,370 All right then I think you guess what we have to do we have to frame the original color image and gray 70 00:05:40,610 --> 00:05:43,060 which is the black and white version of the image. 71 00:05:43,130 --> 00:05:46,700 So we are ready to apply the detect function. 72 00:05:46,700 --> 00:05:48,830 We have the two arguments grand frame. 73 00:05:48,830 --> 00:05:55,200 So basically we're ready to get the original image with the detector rectangles detecting the face and 74 00:05:55,200 --> 00:05:56,100 the eyes. 75 00:05:56,150 --> 00:06:03,200 So we're going to introduce a new voice will here that we're going to call canvas and that will be the 76 00:06:03,200 --> 00:06:11,840 result of the detect function applied to our frame which is the last frame captured by the webcam and 77 00:06:12,140 --> 00:06:14,450 gray the black and white version of this frame. 78 00:06:14,750 --> 00:06:22,720 So what we simply need to do here is to call our detect function and then apply gray and frame gray 79 00:06:23,450 --> 00:06:32,120 and frame and that will get us the last streamed image of the web cam with the rectangle detectors Okay. 80 00:06:32,290 --> 00:06:35,170 Then we need to use some more tricks. 81 00:06:35,170 --> 00:06:41,410 The first week we need to do is to display in an animated way all the successive outputs that is all 82 00:06:41,410 --> 00:06:44,760 the successive processed images in a window. 83 00:06:44,830 --> 00:06:51,450 And to do this we're going to use from RACV to Mudgal the eum show function. 84 00:06:51,710 --> 00:06:54,960 So that's that is just a function that will do that for you. 85 00:06:55,070 --> 00:07:00,080 Display all these processed images in an animated way in a window. 86 00:07:00,170 --> 00:07:02,550 And so this show function takes two arguments. 87 00:07:02,570 --> 00:07:10,910 The first one is video and the second one is the canvas canvas which is the output of the detect function. 88 00:07:10,910 --> 00:07:18,900 That is the original image coming from the webcam but with the detector rectangles right then we need 89 00:07:18,900 --> 00:07:26,070 to use another trick which will be used to stop the webcam and face the election process if we press 90 00:07:26,070 --> 00:07:27,490 on cue on the keyboard. 91 00:07:27,780 --> 00:07:40,910 So the trick to do that is to blame if conditions saying that if see that wait key and you put one and 92 00:07:41,150 --> 00:07:47,320 0 x f f equals equals or in some quotes. 93 00:07:47,420 --> 00:07:49,930 Q That's Q On the keyboard. 94 00:07:50,300 --> 00:07:53,380 So is this then what do we want to do. 95 00:07:53,390 --> 00:07:57,370 Remember where I told you about the while loop the infinite loop. 96 00:07:57,560 --> 00:08:01,680 Well we want to break the process that is set stop everything. 97 00:08:01,790 --> 00:08:07,250 So this well loop will be over and that will allow us to end the process. 98 00:08:07,250 --> 00:08:12,380 So basically it's not very important to understand the techniques and syntax of this. 99 00:08:12,380 --> 00:08:18,500 This just means that if we press on cue on the keyboard Well this will break the process. 100 00:08:18,500 --> 00:08:20,610 No more facial recognition. 101 00:08:21,110 --> 00:08:21,800 OK. 102 00:08:21,860 --> 00:08:22,780 Perfect. 103 00:08:22,790 --> 00:08:31,220 So actually the while loop is over so we can just go back to the main code and we have only two things 104 00:08:31,220 --> 00:08:32,390 to do left. 105 00:08:32,400 --> 00:08:40,460 The first thing is to turn off the webcam and to do this we take our video capture object and we're 106 00:08:40,460 --> 00:08:49,310 going to use this release method to exactly turn up the webcam and eventually we need to destroy the 107 00:08:49,310 --> 00:08:56,300 window inside which all the images were displayed and to do this we take our to Mudgal and then dart 108 00:08:56,690 --> 00:09:06,350 and then we're going to use to destroy all Windows function and there we go the code is over. 109 00:09:06,450 --> 00:09:07,740 Congratulations. 110 00:09:07,740 --> 00:09:11,580 Now we're ready to detect our face and our eyes. 111 00:09:11,580 --> 00:09:14,870 So let's do it right now because everything is ready. 112 00:09:15,120 --> 00:09:15,980 So here we go. 113 00:09:16,020 --> 00:09:17,460 I'm selecting all this. 114 00:09:17,460 --> 00:09:19,080 You have to select all this code. 115 00:09:19,170 --> 00:09:22,270 Make sure that you are in this fuller to be able to load the Cascades. 116 00:09:22,500 --> 00:09:26,020 And if that's the case just press enter. 117 00:09:26,040 --> 00:09:26,790 There we go. 118 00:09:26,820 --> 00:09:31,150 The code is well executed and hi guys. 119 00:09:31,150 --> 00:09:32,320 Welcome to my place. 120 00:09:32,320 --> 00:09:35,510 This is actually where I'm recording all these tutorials for you. 121 00:09:35,510 --> 00:09:37,660 So super happy to have you here. 122 00:09:38,080 --> 00:09:43,300 As you can see well we are detecting very well my face and my eyes. 123 00:09:43,300 --> 00:09:46,580 I think that's the case for you to execute your code. 124 00:09:46,690 --> 00:09:49,120 So of course sometimes we see some false positive. 125 00:09:49,120 --> 00:09:55,030 We can see some rectangle around my mouth detecting the eyes are open to the things that's my eye but 126 00:09:56,110 --> 00:10:02,360 if I don't make too much expression or if I don't talk yet there we go. 127 00:10:02,360 --> 00:10:08,310 We have this blue rectangle detects in my face and these two green rectangles detects in my eyes. 128 00:10:08,450 --> 00:10:09,190 So there we go. 129 00:10:09,200 --> 00:10:10,400 That works pretty well. 130 00:10:10,430 --> 00:10:18,650 If you are not 100 percent convinced of the power of this model Well remember that in the next modules 131 00:10:18,650 --> 00:10:22,570 we'll be using some much more powerful models open. 132 00:10:22,610 --> 00:10:25,810 Definitely doesn't contain the most powerful models. 133 00:10:25,880 --> 00:10:31,750 We actually saw it from module to module 2 will be about object detection and we actually tried Open 134 00:10:31,750 --> 00:10:37,540 survie to detect some very cute bouncing dog that we actually filmed ourself. 135 00:10:37,700 --> 00:10:40,000 But it really turned out to work very bad. 136 00:10:40,160 --> 00:10:44,330 So instead we implemented a much more powerful model based on neural networks. 137 00:10:44,330 --> 00:10:46,050 So hold onto that. 138 00:10:46,070 --> 00:10:49,880 This is going to be interesting and mostly much more powerful. 139 00:10:49,880 --> 00:10:52,330 So I was very happy to show you this. 140 00:10:52,340 --> 00:10:54,170 This was just a warm up. 141 00:10:54,170 --> 00:10:59,270 Keep up the good energy for what's coming next because it's going to be much more challenging but also 142 00:10:59,270 --> 00:11:01,080 much more fascinating. 143 00:11:01,580 --> 00:11:04,100 So I'll see you soon for the next module object detection. 144 00:11:04,100 --> 00:11:06,200 But first I'll give you a little homework. 145 00:11:06,200 --> 00:11:11,860 This will be the homework of this module one face detection and this will be about detecting some of 146 00:11:11,890 --> 00:11:13,510 the features in the face. 147 00:11:13,520 --> 00:11:14,590 So good luck with that. 148 00:11:14,630 --> 00:11:16,610 And until then enjoy computer vision. 15885