Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:00,760 --> 00:00:07,840 Hello and welcome to the practical applications of module three deep convolutional Gants Gannes mean 2 00:00:07,900 --> 00:00:14,110 generative adversarial networks and there are the new big thing in deep learning that I use today to 3 00:00:14,110 --> 00:00:16,000 do some amazing stuff. 4 00:00:16,000 --> 00:00:24,010 One of these things is to generate some fake images based on the vision of real images and that's exactly 5 00:00:24,010 --> 00:00:29,920 what we'll do in this Mudgal and it's part of computer vision because the Gannes will have eyes through 6 00:00:29,920 --> 00:00:37,280 the convolutional layers to watch all these images and then train itself to reproduce some fake images. 7 00:00:37,510 --> 00:00:44,250 But then you have some other amazing applications of gangs that are for example style transfer or face 8 00:00:44,250 --> 00:00:50,680 where you can make a face weap with cycle Gannes basically Gannes other cutting edge Morell's income 9 00:00:50,680 --> 00:00:51,420 television. 10 00:00:51,640 --> 00:00:55,800 So I'm super excited to implement it with you in this Mudgal 3. 11 00:00:55,900 --> 00:01:01,750 And the good news is that this time we will implement the whole model from scratch and this not only 12 00:01:01,750 --> 00:01:07,900 includes the generation of the images that is we're not going to take a pre-trained moralizing module 13 00:01:07,900 --> 00:01:10,630 to to generate some fake images. 14 00:01:10,690 --> 00:01:13,090 We're going to make the whole thing from scratch. 15 00:01:13,090 --> 00:01:17,980 And by that I mean we're going to make a brain we're going to create a brain which will have some eyes 16 00:01:18,280 --> 00:01:22,560 and we're going to train that brain to make it smart and smart. 17 00:01:22,600 --> 00:01:31,000 I mean capable of generating some fake images based on the vision of a lot and lots of real images these 18 00:01:31,090 --> 00:01:38,290 real images will get them from C4 10 which is a very well known data set containing lots and lots of 19 00:01:38,290 --> 00:01:45,130 images that will be the real images on which are deep convolutional Gannes will be trained and based 20 00:01:45,130 --> 00:01:46,130 on its training. 21 00:01:46,180 --> 00:01:51,270 It will then be smart enough to generate some fake images and you're going to see that fake images. 22 00:01:51,370 --> 00:01:55,280 I'm not talking about some random gradients of colors. 23 00:01:55,300 --> 00:01:57,610 It really looks like something real. 24 00:01:57,640 --> 00:02:03,380 So you'll see that in the end we will see that in the final tutorial of this module as usual. 25 00:02:03,530 --> 00:02:07,920 And now if you're ready we're going to start implementing this model from scratch. 26 00:02:08,020 --> 00:02:10,600 The deep convolutional Ganns. 27 00:02:10,930 --> 00:02:17,020 So the first thing we're going to do is open Anaconda because I want to make sure you don't forget to 28 00:02:17,020 --> 00:02:19,260 connect to the virtual platform. 29 00:02:19,270 --> 00:02:20,020 There we go. 30 00:02:20,020 --> 00:02:28,780 So you go to applications on virtual platform and right after it's loaded you click on launch here to 31 00:02:28,780 --> 00:02:33,360 launch the spider ID right. 32 00:02:33,450 --> 00:02:36,180 Spiders coming in here is spider. 33 00:02:36,490 --> 00:02:40,580 OK so now we have to set the right for them as working directory. 34 00:02:40,660 --> 00:02:42,590 So we're going to go to File Explorer. 35 00:02:42,640 --> 00:02:47,560 We're going to go to where your computer vision is that folder it is on my desktop. 36 00:02:47,560 --> 00:02:54,090 We go inside the computer vision it is a folder and now I have to say congratulations you reached module 37 00:02:54,090 --> 00:02:54,680 3. 38 00:02:54,700 --> 00:02:56,320 This is going to be a big module. 39 00:02:56,320 --> 00:03:03,160 We're going to implement a real huge powerful computer vision model from scratch including the training. 40 00:03:03,310 --> 00:03:08,290 So let's do this let's go inside module three and ask for Module 1 and 2. 41 00:03:08,290 --> 00:03:14,440 You have to code this again commented we're going to look at that right now by the way because I really 42 00:03:14,440 --> 00:03:21,580 want you to get the structure of the code and we have our implementation that we'll do ourselves. 43 00:03:21,580 --> 00:03:24,290 It contains more than 100 lines of code. 44 00:03:24,480 --> 00:03:26,710 So that's quite a challenge but we'll make it. 45 00:03:26,830 --> 00:03:32,290 But before we make it I would like to as I just said show you the structure of the code. 46 00:03:32,350 --> 00:03:33,770 So that's the code. 47 00:03:33,790 --> 00:03:39,570 All the lines of the code argumenta and you can have a clear global view of the structure. 48 00:03:39,610 --> 00:03:44,390 It's important sometimes to take a step back and visualize the structure. 49 00:03:44,500 --> 00:03:46,500 That's exactly what we're going to do now. 50 00:03:46,540 --> 00:03:50,230 So we start by importing all the libraries. 51 00:03:50,320 --> 00:03:55,900 So that's almost the same as before except this time we're going to use towards vision to visualize 52 00:03:55,940 --> 00:03:56,970 the images. 53 00:03:57,040 --> 00:04:03,190 Then we set the values of some hyper parameters like the batch size the image size that is. 54 00:04:03,280 --> 00:04:07,540 We set the size of the generated images to 64 by 64. 55 00:04:07,540 --> 00:04:13,300 Then we create some transformations exactly like in module to we'll use some transformations to make 56 00:04:13,300 --> 00:04:18,470 the input images compatible with the neural network of the generator. 57 00:04:18,490 --> 00:04:19,500 Prepare yourself. 58 00:04:19,510 --> 00:04:22,000 This time we're going to have two neural networks. 59 00:04:22,000 --> 00:04:26,810 We're going to have the neural network of the generator and the neural network of the discriminator. 60 00:04:27,000 --> 00:04:32,540 So that's going to be something but these transformations are for the generator. 61 00:04:32,650 --> 00:04:42,360 Then we load the data set from this folder here data that contains the site far 10 data set in batches. 62 00:04:42,490 --> 00:04:45,110 So we just love the data set from this folder. 63 00:04:45,220 --> 00:04:52,720 As you can see here route we take this data folder and then we use torche utils data data loader to 64 00:04:52,720 --> 00:04:56,510 get the images of the data set batch by batch. 65 00:04:56,530 --> 00:05:01,580 Indeed as you can see we specify here the batch size to specify the size of the batch. 66 00:05:01,840 --> 00:05:08,650 And then we use this shuffle equals true here just so that we can get the images in a random order. 67 00:05:08,890 --> 00:05:09,270 All right. 68 00:05:09,280 --> 00:05:10,410 And then none. 69 00:05:10,450 --> 00:05:17,800 Workers equals true means that we're going to have two parallel threads that will load the data and 70 00:05:17,890 --> 00:05:21,010 using the data loader with this shuffle. 71 00:05:21,100 --> 00:05:28,150 And two threads allows us in fact to get the data faster much faster when the data sets are huge which 72 00:05:28,150 --> 00:05:31,510 is the case for C40 and data set. 73 00:05:31,510 --> 00:05:39,430 All right then we define the weights in it function that will take as input a neural network and that 74 00:05:39,430 --> 00:05:42,180 will initialize all the weights of the neural network. 75 00:05:42,250 --> 00:05:47,390 So we will apply the weights and it function to both the neural networks that is the new network of 76 00:05:47,390 --> 00:05:52,670 the generator and a neural network of the discriminator to initialize all the weights. 77 00:05:52,690 --> 00:05:56,060 The right way for the adversarial networks. 78 00:05:56,470 --> 00:05:58,790 And then here comes the structure. 79 00:05:58,840 --> 00:06:03,810 I really want to highlight and then I really want you to have a clear understanding. 80 00:06:03,950 --> 00:06:07,020 That's super important before we tackle this. 81 00:06:07,060 --> 00:06:13,860 So the further we're going to do is defining the generator and we are going to define it through a class 82 00:06:13,870 --> 00:06:16,510 that's the best way to define a neural network. 83 00:06:16,510 --> 00:06:22,000 So we will define the architecture of the new one that work in this first class G which will contain 84 00:06:22,000 --> 00:06:28,700 this architecture and the forward function that will propagate the signal inside this neural network. 85 00:06:29,050 --> 00:06:35,080 Then once we define our generator with this class we'll be able to create the generator itself which 86 00:06:35,080 --> 00:06:38,330 will be an object of the G class. 87 00:06:38,590 --> 00:06:42,620 So that's the first important section of the structure of the code. 88 00:06:42,700 --> 00:06:49,780 Then once we're done with the generator we'll take care of the discriminator and same well-defined architecture 89 00:06:49,780 --> 00:06:51,070 of the discriminator. 90 00:06:51,070 --> 00:06:57,010 With this class D which will contain the architecture itself and same the forward function that will 91 00:06:57,010 --> 00:07:03,130 propagate the signal inside the neural network of the discriminator then once we define all this we'll 92 00:07:03,130 --> 00:07:10,870 be able to create the discriminator itself by creating an object of this previous class declasse and 93 00:07:10,870 --> 00:07:17,560 then that will be a huge step down because we will have created the brain of our computer vision model 94 00:07:17,830 --> 00:07:21,790 because the brain is composed of the generator and the discriminator. 95 00:07:22,000 --> 00:07:25,230 But then we will just have a brain a brain that is not trained yet. 96 00:07:25,240 --> 00:07:30,200 So our hands will still be stupid and therefore we will need to train it. 97 00:07:30,310 --> 00:07:37,540 And that's exactly the last section of the code where we will train the DC Gannes by training two brains 98 00:07:37,630 --> 00:07:43,350 at the same time you know we have this big brain of the DC again that is composed of two brain the brain 99 00:07:43,360 --> 00:07:48,670 there is a neural network or the generator and the other brain the neural network of the discriminator. 100 00:07:48,820 --> 00:07:53,670 So we will train these two brains at the same time according to the process that you saw with curial 101 00:07:53,700 --> 00:07:55,150 in the intuition lectures. 102 00:07:55,240 --> 00:07:59,250 That is you know we first trained a generator with a real image of the data set. 103 00:07:59,320 --> 00:08:05,180 Then we trained the discriminator with a fake image generated before by the generator. 104 00:08:05,320 --> 00:08:08,720 That's the first subset of this big step here. 105 00:08:08,890 --> 00:08:13,660 And then the second subset is two of them the weight of the neural network of the generator. 106 00:08:13,660 --> 00:08:19,750 So we will see that in details great details actually we will code each of these lines of code base 107 00:08:19,750 --> 00:08:26,230 planning what's going on and eventually after we're done with the implementation of the section not 108 00:08:26,230 --> 00:08:32,050 only will we have brains but also these brains will be smart enough to generate some images and since 109 00:08:32,050 --> 00:08:32,970 they'll be smart enough. 110 00:08:33,040 --> 00:08:39,460 Well of course we'll make them generate some fake images and we'll see these fake images in the last 111 00:08:39,460 --> 00:08:40,280 tutorial. 112 00:08:40,450 --> 00:08:41,860 So I can't wait to start. 113 00:08:41,940 --> 00:08:45,410 I'm super excited to code all this with you. 114 00:08:45,490 --> 00:08:47,130 I'm going to close this now. 115 00:08:47,200 --> 00:08:53,320 And here we are back in our non-committed version of the code that is the code where we'll implement 116 00:08:53,320 --> 00:08:54,440 the whole model. 117 00:08:54,700 --> 00:08:59,950 I already prepared the first sections here where we import the library said the hyper parameters create 118 00:08:59,950 --> 00:09:05,440 the transformations load the data set and define the weights in it function so that we can directly 119 00:09:05,570 --> 00:09:10,740 Diven to the implementation of our future smart brains. 120 00:09:11,170 --> 00:09:15,290 So let's do this and getting back into my replow data. 121 00:09:15,360 --> 00:09:22,390 Oh and by the way this last fall the result is an empty folder so far but that will exactly be the folder 122 00:09:22,390 --> 00:09:26,970 that will be populated with the fake images of our decisions. 123 00:09:26,980 --> 00:09:33,170 So here we go let's tackle our deep convolutional Ganns And let's start in the next tutorial. 124 00:09:33,190 --> 00:09:35,130 Until then enjoy computer vision. 14044