Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:00,510 --> 00:00:05,250 Hello and welcome to this new editorial in the British this editorials we described the challenge that 2 00:00:05,250 --> 00:00:05,890 we have. 3 00:00:05,970 --> 00:00:12,750 We have to detect a funny dog on a two seconds video and we will do it through a computer vision based 4 00:00:12,840 --> 00:00:14,730 on deep learning neural networks. 5 00:00:14,790 --> 00:00:18,050 So we already found the right folder now. 6 00:00:18,070 --> 00:00:22,850 This quick to toile I'm going to explain the libraries that we're going to use. 7 00:00:22,890 --> 00:00:27,480 They're already all installed I already prepared the code that will import them. 8 00:00:27,480 --> 00:00:33,000 So there is nothing to do but I think it's important that you understand what we will be using them 9 00:00:33,000 --> 00:00:33,840 for. 10 00:00:33,840 --> 00:00:39,060 So let's start with the first one as you can see the first library when port is torche that's of course 11 00:00:39,150 --> 00:00:47,220 the torch library that contains PI torch which is definitely by far our best weapon to build a new one 12 00:00:47,220 --> 00:00:52,650 that work and do some computer vision and that's for a specific reason it's because by torch content 13 00:00:52,650 --> 00:00:58,890 the dynamic graphs things to which we are able to compute very efficiently the gradients of composition 14 00:00:58,890 --> 00:01:01,590 functions in backward propagation. 15 00:01:01,590 --> 00:01:06,870 You know when we have to update day to wait through stochastic gradient descent Well we have to compute 16 00:01:06,960 --> 00:01:11,380 the gradient of some composition functions because we have several layers. 17 00:01:11,400 --> 00:01:14,240 You know it's a deep neural network so we have several layers. 18 00:01:14,370 --> 00:01:19,740 And therefore it's like one has some functions of the PRI's layer which has some functions of the previous 19 00:01:19,740 --> 00:01:23,160 previous layer so that generates some composition functions. 20 00:01:23,160 --> 00:01:28,140 We have to compute the gradient of these composition functions to have data weights according to how 21 00:01:28,140 --> 00:01:32,980 much they are responsible for the error between the target and the predictions. 22 00:01:33,120 --> 00:01:34,860 So that's where it plays. 23 00:01:35,010 --> 00:01:41,400 And the dynamic graphs is a highly advanced graph structure that allows to have some very fast and efficient 24 00:01:41,610 --> 00:01:43,550 computation of the gradients. 25 00:01:43,560 --> 00:01:49,980 So that's why torture is our first choice then from torche undergrad which is the module responsible 26 00:01:49,980 --> 00:01:51,240 for graden descent. 27 00:01:51,360 --> 00:01:58,710 We are importing the variable class which will be used to convert the tensors into some torche variables 28 00:01:58,710 --> 00:02:02,030 that will contain both the tensor and a gradient. 29 00:02:02,160 --> 00:02:07,710 And then the storage variable containing the tensor in the gradients will be one element of the graph. 30 00:02:07,710 --> 00:02:14,550 Then of course we import CB2 and that even if we're not going to implement a model based on open Hargus 31 00:02:14,580 --> 00:02:20,310 gate we're just importing CB2 because we will be drawing some rectangles around the deck. 32 00:02:20,390 --> 00:02:24,860 But the detection of the dog will not be based on open city Harker's Cate's. 33 00:02:25,020 --> 00:02:33,160 They will be based on as is the neural network that is single shot multa box detection so opens we just 34 00:02:33,180 --> 00:02:41,490 to draw the rectangles then hear from Data Import base transform the classes as label map data is just 35 00:02:41,490 --> 00:02:50,010 a folder that contains the classes based transform and classes then base transform is a class that will 36 00:02:50,010 --> 00:02:56,850 do the required transformations so that the image the input images will be compatible with the neural 37 00:02:56,850 --> 00:02:57,450 network. 38 00:02:57,510 --> 00:03:02,370 You know when we feed the neural network with the input images they have to have a certain format and 39 00:03:02,490 --> 00:03:08,880 base transform will be used to transform the images in this format so that they can be accepted into 40 00:03:08,880 --> 00:03:10,100 the network. 41 00:03:10,380 --> 00:03:12,360 And then what Les's. 42 00:03:12,630 --> 00:03:17,160 Well look Les's is just a dictionary that will do the encoding of the classes. 43 00:03:17,280 --> 00:03:24,390 So for example planes will be encoded as one Dug's will be included as to is just an example it's not 44 00:03:24,540 --> 00:03:26,710 exactly just mapping but that's the idea. 45 00:03:26,730 --> 00:03:31,170 We'll do a mapping because of course we want to work with numbers and not text. 46 00:03:31,230 --> 00:03:36,480 So that's just a very simple dictionary doing the mapping between the text fields of the classes and 47 00:03:36,540 --> 00:03:37,860 some integers. 48 00:03:37,860 --> 00:03:42,400 All right then from the import build SSD. 49 00:03:42,720 --> 00:03:51,600 So first SSD is the library of the single shot multi-book action model and then build as that we import 50 00:03:51,600 --> 00:03:57,120 from the SSD library will be the constructor of the SSD neural network. 51 00:03:57,120 --> 00:04:03,090 And so if you want to have a look you can have a look in this as is digitized and fell to see how it 52 00:04:03,090 --> 00:04:03,680 works. 53 00:04:03,840 --> 00:04:09,300 But it is just a constructor that will build the architecture of this single shot not box detection 54 00:04:09,300 --> 00:04:10,100 model. 55 00:04:10,230 --> 00:04:19,860 And finally image I know is just the library that we'll use to process the images of the video and applying 56 00:04:19,980 --> 00:04:23,250 the detect function that will implement on the images. 57 00:04:23,250 --> 00:04:30,960 So at first we wanted to import pill P L which is another library but image I O actually turns out to 58 00:04:30,960 --> 00:04:34,090 be a much better choice in terms of lines of code. 59 00:04:34,170 --> 00:04:40,350 You'll see we will only have to type two or three lines of code to be able to process the images of 60 00:04:40,350 --> 00:04:41,100 the video. 61 00:04:41,220 --> 00:04:47,430 That is funny Doug and before and apply to detect function that will implement to detect the dog and 62 00:04:47,430 --> 00:04:49,430 the humans on the video. 63 00:04:50,010 --> 00:04:50,370 All right. 64 00:04:50,370 --> 00:04:56,490 So I hope you have now a clear understanding of the libraries that will be used. 65 00:04:56,490 --> 00:04:58,360 It's important to know how they work. 66 00:04:58,560 --> 00:05:05,820 And now with you're going to do is define the detect function that will do the detections. 67 00:05:05,840 --> 00:05:08,650 So let's take a fresh start in the next tutorial. 68 00:05:08,750 --> 00:05:10,790 And so until then enjoy computer vision. 7633