Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:00,330 --> 00:00:06,640 Hello and welcome to the final episode of this module the episode we will get to see the final result. 2 00:00:06,720 --> 00:00:11,790 Not only are we going to see the training of our DC fans but also we're going to see what it's capable 3 00:00:11,790 --> 00:00:13,980 of in terms of art. 4 00:00:14,160 --> 00:00:22,020 We're going to see the generated images of our DC guns in this fall their result here that so far is 5 00:00:22,020 --> 00:00:22,520 empty. 6 00:00:22,620 --> 00:00:29,520 As you can see as far is empty it's going to be populated with all the fake images of our Zygons. 7 00:00:29,520 --> 00:00:31,770 We're going to check if it looks like something. 8 00:00:31,890 --> 00:00:38,370 So if you're ready we'll start by printing some interesting information that we would like to see during 9 00:00:38,370 --> 00:00:39,200 the training. 10 00:00:39,450 --> 00:00:44,300 And that is for example the Epoque how many epochs out of 25. 11 00:00:44,340 --> 00:00:48,870 And also the step that would be nice to see the steps reached because there are actually a lot of steps 12 00:00:48,870 --> 00:00:56,100 in the day loading many batches so let's print of this and mostly what we need to do also is to save 13 00:00:56,520 --> 00:00:59,300 the images in this result's folder. 14 00:00:59,610 --> 00:01:03,570 And then after all this will be good to start the show. 15 00:01:03,720 --> 00:01:07,750 So let's do this let's quickly do those print. 16 00:01:08,040 --> 00:01:10,400 So I'm going to put in quotes. 17 00:01:10,620 --> 00:01:19,680 Well first the box into brackets so I'm going to add a percent D out of percent D because you know this 18 00:01:19,680 --> 00:01:21,980 first percent will correspond to the epoch. 19 00:01:22,230 --> 00:01:24,750 The second percent will correspond to 25. 20 00:01:24,750 --> 00:01:30,430 So you know we'll see like the epical reached out of 25. 21 00:01:30,450 --> 00:01:34,270 Then we're going to do something similar for the steps. 22 00:01:34,290 --> 00:01:37,300 So I'm adding another pair of square brackets. 23 00:01:37,340 --> 00:01:41,500 Percent D out of percent D. 24 00:01:41,510 --> 00:01:49,020 So this time percent here will correspond to the step I and percentage will correspond to the number 25 00:01:49,020 --> 00:01:53,290 of elements and they lower that is Lendale or lower. 26 00:01:53,490 --> 00:01:53,940 All right. 27 00:01:53,940 --> 00:02:02,190 And then after this that is for each step of each epoch we will print the last of the discriminator 28 00:02:02,340 --> 00:02:05,070 which will be here since it's a float. 29 00:02:05,070 --> 00:02:11,970 We're going to add some percent that for f to have a float with four decimals then we're going to add 30 00:02:11,970 --> 00:02:17,080 the same for the loss of the generator. 31 00:02:17,340 --> 00:02:23,170 So I just need to replace the here by less Gee there we go almost over. 32 00:02:23,220 --> 00:02:30,330 Now we need to get out of the quotes at some percent and then in some parenthesis we put the names of 33 00:02:30,330 --> 00:02:36,920 the variables that correspond to these values here the percentages and the percent point for refs. 34 00:02:36,920 --> 00:02:37,680 All right. 35 00:02:37,680 --> 00:02:43,060 So the first value to response to this First percent D is Epoque. 36 00:02:43,080 --> 00:02:50,900 So here we have two input first Epoque then the second percent here corresponds to 25. 37 00:02:51,090 --> 00:02:57,530 So when 25 then the third percent here corresponds to the step. 38 00:02:57,610 --> 00:02:58,890 I see that exactly. 39 00:02:58,890 --> 00:03:07,460 I then the fourth percent here corresponds to the number of mini batches that is Lenn they are lower. 40 00:03:07,530 --> 00:03:12,030 So here I'm adding Lenn data low. 41 00:03:12,330 --> 00:03:19,560 And then finally we have to input the variable names for our two losses corresponding to point for f 42 00:03:19,740 --> 00:03:21,320 and point for f.. 43 00:03:21,330 --> 00:03:27,200 So the first one is the last of the discriminator and the variable for that is. 44 00:03:27,300 --> 00:03:33,510 So I'm adding a comma and the variable for that is of course the R R D. 45 00:03:33,930 --> 00:03:41,250 But then to get the value itself we need to take the data attribute and then in some brackets we add 46 00:03:41,370 --> 00:03:42,030 zero. 47 00:03:42,150 --> 00:03:47,880 That will get us exactly what we want that is the value of the error of the discriminator. 48 00:03:47,880 --> 00:03:49,870 Perfect almost done. 49 00:03:49,890 --> 00:03:52,650 We need to do the same for the generator. 50 00:03:52,830 --> 00:03:58,480 So come out the same and we replace our Ardi by R.G.. 51 00:03:58,670 --> 00:03:59,240 Perfect. 52 00:03:59,280 --> 00:04:01,980 I think our print is ready. 53 00:04:02,280 --> 00:04:02,910 Great. 54 00:04:02,920 --> 00:04:10,950 Now as you can see I would like to save the real images and the generated images of the mini batch every 55 00:04:10,950 --> 00:04:12,490 100 steps. 56 00:04:12,510 --> 00:04:19,760 So now we're going to do is to make an IF condition that will save the images every 100 steps. 57 00:04:19,860 --> 00:04:29,290 And the trick to do that is to an IF condition and then I present 100 equal equal zero. 58 00:04:29,440 --> 00:04:33,000 That's the rest of the division of I buy 100. 59 00:04:33,000 --> 00:04:38,840 So if the rest of the division of I by One hundred is zero that means that I's divided by 100. 60 00:04:38,880 --> 00:04:44,430 And so this way we get this step every 100 steps are right every 100 steps. 61 00:04:44,430 --> 00:04:45,540 What do we do. 62 00:04:45,900 --> 00:04:48,620 Well first we're going to save the real images. 63 00:04:48,900 --> 00:04:56,390 So to do this I'm going to take you tiles which is the shortcut name we gave to torch vision that you 64 00:04:56,940 --> 00:04:59,900 that allowed to save some images with torch vision. 65 00:05:00,190 --> 00:05:03,180 An advanced library of torch for computer vision. 66 00:05:03,460 --> 00:05:10,980 So you teals and then we're going to use to save underscore image function to save the different images 67 00:05:10,990 --> 00:05:15,140 that is the real images on which our model was trained. 68 00:05:15,310 --> 00:05:16,690 And then the fake images. 69 00:05:16,900 --> 00:05:22,810 So we're going to start with the real images and therefore what we have to put here is our batch of 70 00:05:22,840 --> 00:05:25,830 real images that we called real. 71 00:05:25,840 --> 00:05:31,870 That's the first argument and then for the second argument we need to specify in quotes the name of 72 00:05:31,870 --> 00:05:36,300 the path leading to the location where we want to save the real images. 73 00:05:36,580 --> 00:05:45,340 And this path is going to be P.C. S which is a train that will refer to the route then slash results 74 00:05:45,400 --> 00:05:51,260 because the results folder here is a subfolder of our replow the root folder. 75 00:05:51,400 --> 00:05:57,440 So prison S then slash and then after that we need to give the name to the PNAC founders who contend 76 00:05:57,530 --> 00:05:58,620 these images. 77 00:05:58,780 --> 00:06:05,070 And we're going to call them real and this core samples that PNH. 78 00:06:05,080 --> 00:06:13,180 All right and then some percent and in some double quotes we need to specify the string attached to 79 00:06:13,210 --> 00:06:20,440 this percent S and that is that slash results because the name of the folder where we want to save the 80 00:06:20,440 --> 00:06:27,520 images is our results folder and this that here corresponds to the roots that is this working directory 81 00:06:27,520 --> 00:06:28,330 folder. 82 00:06:28,340 --> 00:06:29,120 All right. 83 00:06:29,350 --> 00:06:33,500 And then we can add a third argument which is just to normalize. 84 00:06:33,670 --> 00:06:39,600 And this argument is normalize and we have to set it equal to true. 85 00:06:40,090 --> 00:06:40,840 Perfect. 86 00:06:40,870 --> 00:06:48,430 So we saved our batch of real images contained in real here and now we're going to do the same for the 87 00:06:48,430 --> 00:06:49,550 fake images. 88 00:06:49,600 --> 00:06:59,080 So we're going to get these fake images by calling our next G network again to which we feed the noise 89 00:06:59,170 --> 00:07:00,330 random vector. 90 00:07:00,550 --> 00:07:03,400 So we're just doing it again to get the fake images. 91 00:07:03,400 --> 00:07:10,720 And now we can save them so I'm just going to copy this line pasted right below. 92 00:07:10,790 --> 00:07:14,300 And now I just need to replace a few things first. 93 00:07:14,330 --> 00:07:22,170 This time we're not setting the real image but fake that data that's what contains the fake images. 94 00:07:22,170 --> 00:07:23,620 Then here we keep present. 95 00:07:23,630 --> 00:07:31,010 As for the name of the path leading to the folder where we save our images but then we're not going 96 00:07:31,010 --> 00:07:32,000 to call them this time. 97 00:07:32,030 --> 00:07:40,130 Real samples but fake underscores samples underscore airpark underscore. 98 00:07:40,220 --> 00:07:45,200 And here I'm going to put the number of the epoch when the fake images are saved. 99 00:07:45,200 --> 00:07:51,540 And to do this I'm going to specify here a double with three integers so present. 100 00:07:51,570 --> 00:07:52,450 Oh 3D. 101 00:07:52,610 --> 00:07:53,810 And then that PNH. 102 00:07:54,050 --> 00:07:54,410 All right. 103 00:07:54,410 --> 00:08:04,040 And then since I added a new variable here with the percent of 3D then besides the path of the results 104 00:08:04,100 --> 00:08:10,950 in double quotes I need to add the variable corresponding to 0 3 and according to you what is it. 105 00:08:11,120 --> 00:08:13,280 Well that's of course the epoch. 106 00:08:13,610 --> 00:08:19,660 So that's where you will get the fake images and we'll know from which book they are coming. 107 00:08:19,660 --> 00:08:20,160 All right. 108 00:08:20,260 --> 00:08:23,850 We will know in which epoch they were produced by the generator. 109 00:08:23,900 --> 00:08:24,380 Perfect. 110 00:08:24,410 --> 00:08:26,900 And then I'm keeping normalize equal. 111 00:08:27,190 --> 00:08:27,830 All right. 112 00:08:27,920 --> 00:08:30,480 And now we're ready to watch the final result. 113 00:08:30,480 --> 00:08:35,070 And so if you're ready now it's time for the show. 114 00:08:35,130 --> 00:08:35,930 And there we go. 115 00:08:35,930 --> 00:08:37,300 I just executed. 116 00:08:37,340 --> 00:08:42,130 I selected all the code and pressed command or control plus enter. 117 00:08:42,170 --> 00:08:43,380 And there we go. 118 00:08:43,460 --> 00:08:45,640 We have started the training. 119 00:08:45,880 --> 00:08:52,200 So as you can see this is the first epoch and the first steps with 0 1 2 3 4. 120 00:08:52,430 --> 00:08:57,950 And for each of the step in each book we get Indeed the last of the discriminator and the last of the 121 00:08:57,950 --> 00:08:59,140 generator. 122 00:08:59,150 --> 00:09:04,950 So now it's going to take a while it's actually going to take several hours on my computer. 123 00:09:05,060 --> 00:09:09,200 I'm going to let my computer do all this work for me. 124 00:09:09,200 --> 00:09:10,880 We're not going to watch the whole training. 125 00:09:11,150 --> 00:09:17,650 And at the end of the training I'll see you back and we will see the fake images generated by our decisions 126 00:09:17,960 --> 00:09:19,830 and we'll see if it looks like something. 127 00:09:20,010 --> 00:09:24,530 So let's let this run and I'll see you at the end of the training. 128 00:09:24,530 --> 00:09:27,130 All right the training is over. 129 00:09:27,170 --> 00:09:29,240 It actually took more time than I thought. 130 00:09:29,240 --> 00:09:33,230 When I woke up after a long night of sleep well it was still not over. 131 00:09:33,230 --> 00:09:37,000 So I guess it took more than eight hours in my computer. 132 00:09:37,010 --> 00:09:39,910 Indeed I don't sleep any more three hours per day. 133 00:09:39,920 --> 00:09:46,520 I noticed it was bad for life expectancy and I still want to be able to make some courses for your children 134 00:09:46,610 --> 00:09:47,730 and grandchildren. 135 00:09:47,870 --> 00:09:51,860 But as long as we have the final results that's all good. 136 00:09:51,860 --> 00:09:57,200 So I'm going to show them to you now and we'll see if we can call our deep convolutional dance. 137 00:09:57,260 --> 00:09:58,890 A great artist. 138 00:09:58,910 --> 00:09:59,210 All right. 139 00:09:59,210 --> 00:10:07,460 So before I show you the first samples I would like to show you the real samples just to see on which 140 00:10:07,580 --> 00:10:13,550 images real images are a computer vision model learned to generate some fake images. 141 00:10:13,550 --> 00:10:14,980 So these are the images. 142 00:10:15,140 --> 00:10:15,810 All good. 143 00:10:15,830 --> 00:10:20,450 Now let's see what it was capable of creating. 144 00:10:20,450 --> 00:10:22,700 All right so let's start with the first samples. 145 00:10:22,880 --> 00:10:24,820 Nothing special here. 146 00:10:24,860 --> 00:10:27,150 We cannot call it art at all. 147 00:10:27,320 --> 00:10:28,840 But then what about the second one. 148 00:10:28,970 --> 00:10:35,210 The second one is already better the second one looks more like something but still it still looks like 149 00:10:35,210 --> 00:10:42,020 some kind of smoke except for this one maybe that looks like a mountain but I think we'll get better 150 00:10:42,020 --> 00:10:42,720 than that. 151 00:10:43,550 --> 00:10:45,020 Then some pools. 152 00:10:45,150 --> 00:10:46,080 Number two. 153 00:10:46,080 --> 00:10:54,890 So the numbers here correspond to the box that was given at about 0 1 2 3 until about 24 25. 154 00:10:55,050 --> 00:10:56,230 Back in Seoul. 155 00:10:56,370 --> 00:10:58,910 And this one this one is actually pretty good. 156 00:10:59,190 --> 00:11:01,800 We start to see something here. 157 00:11:01,830 --> 00:11:07,230 We still need a bit of imagination to figure out what's inside the image here. 158 00:11:07,230 --> 00:11:12,580 I see him for example to see some kind of a duck on a on the sea or the ocean. 159 00:11:12,600 --> 00:11:14,160 I don't know if you see the same thing. 160 00:11:14,280 --> 00:11:20,270 Well maybe my imagination is playing some tricks but I can start to see something here. 161 00:11:20,280 --> 00:11:23,130 Let's look at the third fake samples. 162 00:11:23,300 --> 00:11:28,570 So the fake images of the third book are right definitely better here. 163 00:11:28,590 --> 00:11:29,900 Here we can see a human. 164 00:11:29,940 --> 00:11:31,410 I guess it's a human here. 165 00:11:31,440 --> 00:11:33,000 I think I see a squirrel. 166 00:11:33,030 --> 00:11:34,280 I don't know if you agree. 167 00:11:34,530 --> 00:11:34,950 OK. 168 00:11:34,950 --> 00:11:36,020 Much better. 169 00:11:36,240 --> 00:11:40,230 Let's look at the other ones much better here as well. 170 00:11:40,230 --> 00:11:40,840 All right. 171 00:11:40,860 --> 00:11:43,420 Let's look at number six. 172 00:11:43,530 --> 00:11:44,880 Much better here. 173 00:11:45,030 --> 00:11:46,760 We can see it better and better each time. 174 00:11:46,770 --> 00:11:52,620 I don't know if you agree but to me the fake images really start to look like some real images even 175 00:11:52,620 --> 00:11:54,930 if it's not perfectly net. 176 00:11:55,200 --> 00:11:56,590 It's still a little bit blurry. 177 00:11:56,610 --> 00:11:59,880 But still we can see some objects here. 178 00:11:59,880 --> 00:12:00,170 All right. 179 00:12:00,180 --> 00:12:03,740 Let's look at number 10 for example to see if it's already much better. 180 00:12:04,080 --> 00:12:06,120 Yes it looks pretty good. 181 00:12:06,330 --> 00:12:09,050 Let's have a look at number 15. 182 00:12:10,840 --> 00:12:16,480 Still very good so perhaps you didn't need that many a book perhaps you could try to do the training 183 00:12:16,480 --> 00:12:18,970 with five or up to 10 bucks. 184 00:12:19,170 --> 00:12:23,940 But definitely after a certain number of book we see some great results. 185 00:12:24,130 --> 00:12:27,720 And here it's even more visible than before even more clear. 186 00:12:27,910 --> 00:12:29,260 And these are pretty cool images. 187 00:12:29,260 --> 00:12:31,010 We can see some nice colors now. 188 00:12:31,010 --> 00:12:32,270 So I would. 189 00:12:32,270 --> 00:12:38,950 They're calling this model an artist maybe not Picasso but definitely a better artist than me. 190 00:12:38,950 --> 00:12:40,010 So that's pretty cool. 191 00:12:40,060 --> 00:12:46,960 And that's actually the end of this Mudgal congratulations for having implemented the deep convolutional 192 00:12:46,960 --> 00:12:47,600 Ganns. 193 00:12:47,680 --> 00:12:49,660 That was definitely not good stuff. 194 00:12:49,680 --> 00:12:54,150 You coded almost 150 lines of code so well done. 195 00:12:54,280 --> 00:12:55,170 Awesome job. 196 00:12:55,210 --> 00:13:01,550 You not only implemented the deep convolutional Ganns But also you smashed the three modules of this 197 00:13:01,550 --> 00:13:02,180 course. 198 00:13:02,240 --> 00:13:06,300 Well keep in mind you implemented some cutting edge models. 199 00:13:06,430 --> 00:13:11,700 I remind that SSD is the state of the art moral and object detection. 200 00:13:11,710 --> 00:13:14,650 It beats the faster our CNN and you know. 201 00:13:14,650 --> 00:13:21,940 So at the time I'm speaking and I hope this course will last but at the time I'm speaking you did implement 202 00:13:22,120 --> 00:13:25,780 a state of the art model in computer vision and deep learning. 203 00:13:25,780 --> 00:13:28,850 So really really really you can be proud of yourself. 204 00:13:29,020 --> 00:13:35,560 Keep up that great work and great passion for the coming courses and some more modules. 205 00:13:35,590 --> 00:13:38,170 This adventure is definitely not over. 206 00:13:38,170 --> 00:13:39,720 We will learn so much more. 207 00:13:39,750 --> 00:13:45,540 We are dedicated instructors really happy to share our knowledge so there will be more. 208 00:13:45,670 --> 00:13:47,750 And until then enjoy machine learning. 209 00:13:47,770 --> 00:13:49,320 Enjoy the journey enjoy. 210 00:13:49,390 --> 00:13:51,640 I mostly enjoy computer vision. 20348