Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:00,560 --> 00:00:07,190 Hello and welcome to the practical applications of module to object detection I'm super excited to start 2 00:00:07,190 --> 00:00:08,830 this module for two reasons. 3 00:00:08,840 --> 00:00:12,260 First one is we are taking things at the next level now. 4 00:00:12,440 --> 00:00:18,560 As I told you open TV is not the most powerful model and the model we will implement in this module 5 00:00:18,710 --> 00:00:24,410 is much more powerful because it is based on deep learning and neural networks that computer vision 6 00:00:24,500 --> 00:00:26,050 where the computer will have a brain. 7 00:00:26,150 --> 00:00:27,580 That's exactly what it means. 8 00:00:27,590 --> 00:00:30,890 And the second reason is that we have an exciting challenge. 9 00:00:30,890 --> 00:00:34,830 I will show you a video of a very cute dog bouncing on the field. 10 00:00:35,000 --> 00:00:42,800 And our challenge will be to detect the dog will be to implement some program that will detect the dog 11 00:00:42,890 --> 00:00:43,600 in the video. 12 00:00:43,790 --> 00:00:49,400 So it's good that you see several ways of doing some computer vision in the first module you learn how 13 00:00:49,400 --> 00:00:53,240 to do some face detection through a webcam. 14 00:00:53,450 --> 00:00:58,160 And now you're going to learn how to do some object detection on a video directly. 15 00:00:58,160 --> 00:01:05,280 Now before we start I would like to say a big thank you to this developer here next to Groote. 16 00:01:05,360 --> 00:01:07,390 That's a picture of him in a horseshoe. 17 00:01:07,580 --> 00:01:13,850 He's the creator of the PI torch implementation of single shot multi-book detector that we're going 18 00:01:13,850 --> 00:01:15,420 to use in this module. 19 00:01:15,440 --> 00:01:19,850 So thank you very much for sharing this and make it open source. 20 00:01:19,850 --> 00:01:27,710 We actually tried several object detection models we tried the first are CNN the yellow open CD and 21 00:01:27,860 --> 00:01:33,320 the SSD and we obtain the best result with the single shot multi-book detection. 22 00:01:33,320 --> 00:01:38,750 Not only we obtain the best result with this moral but also if you look at the paper you will see that 23 00:01:38,810 --> 00:01:45,270 on the tested cases the single shot multiplexed detection model beats yolo and fester are CNN. 24 00:01:45,440 --> 00:01:53,370 So that's why our choice for the ultimate objective texture model of muchall 2 was single shot multi 25 00:01:53,380 --> 00:01:58,870 box detection and the best implementation we found was from this developer Max agreed. 26 00:01:58,990 --> 00:02:00,230 So thank you so much. 27 00:02:00,230 --> 00:02:06,020 Thank you for sharing this pre-trained moral it's actually a pre-trained moral that was trained to detect 28 00:02:06,110 --> 00:02:13,790 between 30 and 40 objects including cars dogs horses ships boats planes and more. 29 00:02:13,850 --> 00:02:18,410 So a very useful not all that you could use for your own business problems. 30 00:02:18,410 --> 00:02:24,020 We're going to go inside the SSD in this module and we're going to learn how to use it and how to detect 31 00:02:24,110 --> 00:02:26,610 any object on any video. 32 00:02:26,660 --> 00:02:28,650 So that's going to be a pretty exciting module. 33 00:02:28,670 --> 00:02:30,640 I can't wait to show you this video of this Doug. 34 00:02:30,650 --> 00:02:35,860 I really like this Doug it's actually Carol who filmed this little dog with a drone. 35 00:02:36,020 --> 00:02:41,240 So the first thing we're going to do now is we're going to open Anaconda because I want to make sure 36 00:02:41,540 --> 00:02:44,900 that you don't forget to connect to the virtual platform. 37 00:02:45,140 --> 00:02:46,520 So let's do it. 38 00:02:46,550 --> 00:02:49,130 I'm opening an icon the Navigator. 39 00:02:49,130 --> 00:02:51,430 You have to find an X on the Navigator. 40 00:02:51,470 --> 00:02:57,440 If you're on Windows you will find it in the list of programs and on Linux you can open it through either 41 00:02:57,460 --> 00:03:00,410 to terminal or in the programs. 42 00:03:00,410 --> 00:03:03,490 All right so now and I can the navigator is opened. 43 00:03:03,650 --> 00:03:08,510 And don't forget to do this applications on virtual platform. 44 00:03:08,510 --> 00:03:09,020 There we go. 45 00:03:09,020 --> 00:03:11,670 Now we're connected to the virtual platform environment. 46 00:03:11,780 --> 00:03:14,330 And so we were ready to launch spider. 47 00:03:14,540 --> 00:03:17,010 And we don't have to install anything. 48 00:03:17,030 --> 00:03:19,760 Everything is already installed on the virtual platform. 49 00:03:19,850 --> 00:03:23,900 So we are ready to execute the code and I'm super happy to start. 50 00:03:24,320 --> 00:03:30,530 But before we start implementing the code we have to be in the right folder because there are some external 51 00:03:30,530 --> 00:03:34,230 files that we'll be calling when executing the code in the end. 52 00:03:34,400 --> 00:03:38,880 So anyway we always have to be in the right folder where we implement the code. 53 00:03:38,930 --> 00:03:43,040 So that's the first thing I'm going to do now I'm going to go to my desktop. 54 00:03:43,130 --> 00:03:46,500 This is where my computer vision is at full that is. 55 00:03:46,670 --> 00:03:53,870 So let's double click on it and now congratulations you reached module to object detection. 56 00:03:53,870 --> 00:03:54,280 All right. 57 00:03:54,290 --> 00:03:55,330 That's the folder. 58 00:03:55,460 --> 00:03:58,190 Let's quickly describe what's inside this folder. 59 00:03:58,250 --> 00:04:04,880 So you have data is just a folder that contains the classes based transform that will do the required 60 00:04:04,880 --> 00:04:11,300 transformations so that the input images will be compatible with the neural network then. 61 00:04:11,540 --> 00:04:15,800 Funny Doug is of course this video of this very funny. 62 00:04:15,810 --> 00:04:17,590 We will be trying to detect. 63 00:04:17,660 --> 00:04:19,420 I will show you this video in a second. 64 00:04:19,670 --> 00:04:25,990 But then layer's is another folder that contained some other tools for the detection and the multi-book 65 00:04:26,020 --> 00:04:28,270 as part of the SSD. 66 00:04:28,280 --> 00:04:35,480 Then you have of course to code the commented version of the code object detection commented where you 67 00:04:35,480 --> 00:04:41,300 have the whole code that will implemented this module come into line by line so that can be useful either 68 00:04:41,300 --> 00:04:42,770 before or after. 69 00:04:42,770 --> 00:04:48,350 Actually I also recommend to have a look at this before so that you can expect what you need to understand 70 00:04:48,500 --> 00:04:55,100 and therefore when I explain it you might understand it more easily than this code object detection 71 00:04:55,520 --> 00:04:57,540 is actually going to open it. 72 00:04:57,650 --> 00:05:03,280 Is the code that we will implement in this module so I already imported the libraries. 73 00:05:03,280 --> 00:05:05,620 I'm going to describe what those libraries are. 74 00:05:05,620 --> 00:05:11,440 But anyway this is where I will implement this whole code and when I'm done implementing it with you 75 00:05:11,800 --> 00:05:14,510 I will rename it object detection. 76 00:05:14,530 --> 00:05:17,880 No comment that you can have the commented version of the code. 77 00:05:17,950 --> 00:05:22,220 And the non-committed version of the code you can practice to recoated. 78 00:05:22,390 --> 00:05:24,350 That's excellent practice. 79 00:05:24,400 --> 00:05:31,400 Then we have the SSD that you wife file which contains the architecture of the single shot multi-button 80 00:05:31,400 --> 00:05:32,420 action model. 81 00:05:32,440 --> 00:05:38,590 We won't implement this one because I want it to keep what's most important for you to understand in 82 00:05:38,890 --> 00:05:40,990 this object detection implementation. 83 00:05:41,170 --> 00:05:45,290 Because if we implement the whole model this will be overwhelming. 84 00:05:45,400 --> 00:05:48,700 And you might miss what's at the heart of the model. 85 00:05:48,760 --> 00:05:51,030 So I prefer to proceed this way. 86 00:05:51,100 --> 00:05:54,020 And this model is all the architecture. 87 00:05:54,070 --> 00:05:59,620 And in fact after you watched in tuition lectures you will be totally able to understand what's going 88 00:05:59,620 --> 00:06:00,010 on. 89 00:06:00,060 --> 00:06:04,880 Well it's mostly about the architecture with all the boxes how they're defined. 90 00:06:05,020 --> 00:06:10,580 But then the heart of the model will be in this implementation objective section. 91 00:06:11,050 --> 00:06:19,150 And then finally this file is the file we will be loading to get the pre-trained SS DeMaio and more 92 00:06:19,150 --> 00:06:26,690 precisely this is the file that contains the weight of the SSD neural network that was already pre-trained. 93 00:06:26,890 --> 00:06:33,340 So we will be loading this file with torch the torch library that load which is a function of torche 94 00:06:33,840 --> 00:06:39,610 this tortured load function will open a tensor a tensor that will contain the weight of this already 95 00:06:39,610 --> 00:06:46,240 pre-trained neural network and then through a mapping with a dictionary we will transfer these weights 96 00:06:46,510 --> 00:06:48,370 to the model we implement. 97 00:06:48,550 --> 00:06:53,890 So basically this just contains the weight of an already pre-trained model and we will transfer these 98 00:06:53,890 --> 00:06:56,880 weights to the model we will implement. 99 00:06:56,890 --> 00:06:58,820 I hope that's clear and that's it. 100 00:06:58,930 --> 00:07:00,620 So I guess we're ready to start. 101 00:07:00,640 --> 00:07:05,140 And therefore let's start with some funny video of this very cute Doug. 102 00:07:05,230 --> 00:07:08,350 So I'm going to double click on the video. 103 00:07:08,350 --> 00:07:08,920 There we go. 104 00:07:08,920 --> 00:07:11,220 That's the video you can recognize. 105 00:07:11,220 --> 00:07:14,870 Kiril going to put that at the beginning. 106 00:07:15,030 --> 00:07:15,980 So this is curial. 107 00:07:15,990 --> 00:07:17,120 This is the dog. 108 00:07:17,190 --> 00:07:20,580 This video last two seconds so that it doesn't take too much time. 109 00:07:20,670 --> 00:07:22,340 When you try to Marans video. 110 00:07:22,470 --> 00:07:25,980 But we will totally have time to see the dog bouncing. 111 00:07:25,980 --> 00:07:27,000 It's very funny. 112 00:07:27,000 --> 00:07:27,770 Check this out. 113 00:07:30,670 --> 00:07:35,450 It you Doug when I watched this Doug I absolutely want to play with him. 114 00:07:37,320 --> 00:07:40,570 And actually you can see Kyrle piloting the drone behind. 115 00:07:40,770 --> 00:07:42,280 So there we go. 116 00:07:42,300 --> 00:07:43,400 That's the video. 117 00:07:43,520 --> 00:07:50,700 And actually the model we will implement will not only detect the dog bouncing on the field but also 118 00:07:50,700 --> 00:07:51,930 this human here. 119 00:07:51,930 --> 00:07:57,180 And you will see that it will also manage to detect curial even if you're really far actually from the 120 00:07:57,180 --> 00:07:57,800 video. 121 00:07:58,050 --> 00:08:04,440 And I'd like to tell you now that actually you know for you it's very easy to detect the drug but the 122 00:08:04,440 --> 00:08:06,720 drug is actually pretty small in the video. 123 00:08:06,750 --> 00:08:08,540 You know it's a pretty small object. 124 00:08:08,730 --> 00:08:14,220 And when we tried to detect that with open city we had extremely bad results. 125 00:08:14,220 --> 00:08:19,160 It couldn't detect the drug it couldn't detect what it was and there were some rectangles everywhere 126 00:08:19,170 --> 00:08:20,830 you can actually try yourself. 127 00:08:21,030 --> 00:08:26,580 And that's why I wanted to highlight that open Svea is definitely not among the most powerful models 128 00:08:26,850 --> 00:08:32,970 but you'll see that the more we will implement in this module will do a perfect job at detecting this 129 00:08:32,970 --> 00:08:39,120 drug even if it is small and even if there is not a perfect contrast between the drug and environment 130 00:08:39,120 --> 00:08:42,400 you know it's not like we have a white environment with a black dog. 131 00:08:42,630 --> 00:08:44,920 The dog can be confused with something else. 132 00:08:45,120 --> 00:08:54,450 So you'll be convinced of the power of this model at the end and I can't wait to show you how this model 133 00:08:54,540 --> 00:08:57,180 is going to do all right. 134 00:08:57,230 --> 00:08:58,570 That's what I wanted to catch. 135 00:08:58,570 --> 00:09:02,880 You know sometimes it really doesn't look like a dyke but you'll see what happens. 136 00:09:02,890 --> 00:09:09,220 Let's implement the SSD single shot multi-book detection and let's do that from the next tutorial. 137 00:09:09,280 --> 00:09:11,320 Until then enjoy computer vision. 14347