Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:00,270 --> 00:00:06,240 Hello and welcome back to module one face recognition you had your intuition lectures with Curiel to 2 00:00:06,240 --> 00:00:08,040 understand how Cascades work. 3 00:00:08,190 --> 00:00:13,830 So now we're going to put them into action and we're going to put them into action to do some face recognition 4 00:00:14,190 --> 00:00:19,440 and we're going to do it through the web cam that is that I'm going to put on my web cam and we're going 5 00:00:19,440 --> 00:00:26,310 to try to protect my face and my eyes and some extra stuff that will be for the homework but that's 6 00:00:26,310 --> 00:00:32,310 what we'll be doing we'll be working with the Cascades that is with open city to do this face recognition. 7 00:00:32,340 --> 00:00:38,790 So as you can see I launched my spider in my virtual platform environment which is a virtual environment. 8 00:00:38,790 --> 00:00:44,010 And now the first thing we have to do when launching spider is to find the right folder because this 9 00:00:44,010 --> 00:00:47,160 folder is going to contain the Cascades ex-MIL. 10 00:00:47,190 --> 00:00:52,090 And so we need to be in this folder because we will be calling the Cascades to execute our code. 11 00:00:52,290 --> 00:00:57,420 So let's do that right now we're going to go to File Explorer and we're going to go to our computer 12 00:00:57,420 --> 00:01:01,130 vision A.D.C. folder which is right now on my desktop. 13 00:01:01,320 --> 00:01:06,870 So you go into this folder and you go to module one face recognition the first module of discours. 14 00:01:06,870 --> 00:01:08,970 So I'm very excited to start this. 15 00:01:09,270 --> 00:01:14,390 And actually right now I want to tell you that the more we evolve in the course the more the models 16 00:01:14,400 --> 00:01:21,150 will be working with will be powerful and that means that actually open city is great but it's definitely 17 00:01:21,150 --> 00:01:26,730 not the most powerful models of discourse and of the virtual platform that some of you might have noticed 18 00:01:26,730 --> 00:01:29,470 that it is no longer in development. 19 00:01:29,490 --> 00:01:34,740 We have some much better models today we will see that with object detection for object detection we'll 20 00:01:34,770 --> 00:01:37,870 be using something much more powerful than open city. 21 00:01:37,870 --> 00:01:42,690 I will let you find out about the surprise but still open city is going to do some great stuff here 22 00:01:42,690 --> 00:01:44,150 for face recognition. 23 00:01:44,160 --> 00:01:49,260 But keep in mind that you should really follow the course until the end because we will be working with 24 00:01:49,260 --> 00:01:51,660 some much more powerful models. 25 00:01:51,690 --> 00:01:53,730 But let's start with open city. 26 00:01:53,760 --> 00:01:55,140 So this is the right folder. 27 00:01:55,170 --> 00:01:57,910 Make sure that you're in the right folder. 28 00:01:58,170 --> 00:02:01,290 And now I'm quickly going to describe the folder. 29 00:02:01,470 --> 00:02:05,170 So we have four files. 30 00:02:05,280 --> 00:02:11,280 The first file is the code that we're about to implement right now but the connected version of the 31 00:02:11,280 --> 00:02:11,610 code. 32 00:02:11,610 --> 00:02:15,620 So if you open this file you will have each line of the code commented. 33 00:02:15,720 --> 00:02:18,410 So that's in case you don't have time to watch the video. 34 00:02:18,480 --> 00:02:23,010 But really I strongly recommend to watch the video because I will be seeing much more than the comments 35 00:02:23,420 --> 00:02:28,440 and this is actually this face recognition that bites and fowl. 36 00:02:28,530 --> 00:02:31,250 And this is actually where we're going to implement the code. 37 00:02:31,350 --> 00:02:33,890 Then we have the cascade for the eye. 38 00:02:34,000 --> 00:02:37,900 That's the cascade that we'll use to detect the eyes and the web. 39 00:02:37,920 --> 00:02:38,700 That is my eyes. 40 00:02:38,700 --> 00:02:41,750 But you will use the code to detect your eyes too. 41 00:02:42,060 --> 00:02:44,770 And the cascade to detect the face. 42 00:02:44,880 --> 00:02:48,090 So these are X-amount files that's not very important. 43 00:02:48,090 --> 00:02:53,120 But this is what we'll use from open city to do these face protections. 44 00:02:53,130 --> 00:02:55,020 So there we go that's the folder. 45 00:02:55,020 --> 00:02:56,700 Make sure that you have all these files. 46 00:02:56,730 --> 00:02:57,900 And now we're going to start. 47 00:02:57,930 --> 00:03:03,120 So as you can see there is this first code section here in the libraries. 48 00:03:03,210 --> 00:03:10,110 And of course the only library that we need here is open city and that is silly to say Make sure you 49 00:03:10,110 --> 00:03:11,260 import too. 50 00:03:11,280 --> 00:03:17,770 I'm going to execute the whole code in the end but make sure that you import 3:2 that's entirely compulsory. 51 00:03:18,180 --> 00:03:24,280 And the reason why we don't import much more is that actually we're not doing some deep learning here. 52 00:03:24,360 --> 00:03:30,000 If you noticed in the intuition lectures the Cascades are actually not some neural networks. 53 00:03:30,000 --> 00:03:37,110 There is a deep series of filters but this is not a series of layers so open city cascades are not based 54 00:03:37,200 --> 00:03:38,460 on any neural networks. 55 00:03:38,670 --> 00:03:44,460 And that's why for open city we don't need to import any neural network libraries like the torched libraries 56 00:03:44,460 --> 00:03:46,180 or tenso flow or anything. 57 00:03:46,260 --> 00:03:48,240 We just need open city. 58 00:03:48,480 --> 00:03:51,800 And what we need from open city is mainly the Cascades. 59 00:03:51,990 --> 00:03:54,470 And that is exactly what we're going to do now. 60 00:03:54,570 --> 00:03:59,910 We're going to be loading the Cascades one cascade for the face and one cascade for the eyes and to 61 00:03:59,910 --> 00:04:06,090 do this we're going to create two objects because these are going to be some objects of some class that 62 00:04:06,090 --> 00:04:08,920 is called Cascade classifier. 63 00:04:09,240 --> 00:04:15,890 And this class will take as arguments exactly the SML file that contains the cascade. 64 00:04:16,080 --> 00:04:21,990 So the first object will take an argument while the frontal face cascade and the second object will 65 00:04:21,990 --> 00:04:22,910 take its argument. 66 00:04:22,950 --> 00:04:24,620 I Harker's Kate. 67 00:04:24,930 --> 00:04:26,170 So let's do this. 68 00:04:26,160 --> 00:04:33,750 We're going to create a first object here for the face and we're going to call it face cascade. 69 00:04:33,750 --> 00:04:35,210 All right so that's the object. 70 00:04:35,230 --> 00:04:39,660 And now since it's an object of some class we need to call the class. 71 00:04:39,660 --> 00:04:45,950 Now this class is actually a class of the CB2 module from the open library. 72 00:04:46,230 --> 00:04:47,170 So what. 73 00:04:47,340 --> 00:04:56,600 And here we type the name of the class which is cascade class C fire cascade classifier. 74 00:04:56,790 --> 00:04:57,270 All right. 75 00:04:57,300 --> 00:05:03,530 And as I just said in argument here we just need to put one argument which is the name of the ex-MIL 76 00:05:03,530 --> 00:05:04,620 Harker's Kate. 77 00:05:04,820 --> 00:05:11,690 And since we're starting with a face well we need to put exactly that heart cascade underscore frontal 78 00:05:11,690 --> 00:05:14,080 face underscore default. 79 00:05:14,120 --> 00:05:18,290 SNL but we need to put that in quotes. 80 00:05:18,290 --> 00:05:19,200 There we go. 81 00:05:19,340 --> 00:05:29,900 Let's do that har cascade underscore frontal face underscore default and there we go. 82 00:05:29,900 --> 00:05:34,700 That SNL and that's will import the cascade for the face. 83 00:05:34,810 --> 00:05:41,660 So as a reminder the cascade is a series of filters that will apply one after the other to detect the 84 00:05:41,660 --> 00:05:42,770 face. 85 00:05:42,770 --> 00:05:47,750 All right so now we're going to do exactly the same for the eye. 86 00:05:47,960 --> 00:05:55,940 So I'm copying this line to be efficient and basic in here and I'm just going to play face Cascades 87 00:05:56,000 --> 00:05:58,040 by I cascade. 88 00:05:58,300 --> 00:05:59,120 All right. 89 00:05:59,180 --> 00:06:06,080 And of course here I need to replace Hargus Cade's frontal face default SNL by hardcase gate. 90 00:06:06,260 --> 00:06:10,340 I am just selecting this and replacing it by. 91 00:06:10,370 --> 00:06:11,440 There we go. 92 00:06:11,450 --> 00:06:17,720 So now we have our two Cascades the cascades of the face and the cascade of the eye so perfect we were 93 00:06:17,720 --> 00:06:23,540 ready to move onto the next code section which will be about implementing a function that will do the 94 00:06:23,540 --> 00:06:28,760 detection itself that will use those cascades to detect the face and the eye. 95 00:06:28,940 --> 00:06:33,740 But you're going to see that there is a little trick which is about computational efficiency and that 96 00:06:33,740 --> 00:06:38,960 is we're going to do the phase detection on the global referential the main differential but we're going 97 00:06:38,960 --> 00:06:44,630 to do the detection and the referential of the face and that will save us some computation time. 98 00:06:44,630 --> 00:06:47,360 All right so let's do that in the next tutorial. 99 00:06:47,390 --> 00:06:49,190 And until then the computer vision. 10497