1 00:00:00,810 --> 00:00:01,170 Hello. 2 00:00:01,200 --> 00:00:01,770 Welcome back. 3 00:00:01,770 --> 00:00:09,000 In this lesson we are going to start developing a war cut or no cut neuronal network in Python. 4 00:00:09,280 --> 00:00:15,200 Um one thing you would realize is when deploying machine learning on embedded devices what you wanna 5 00:00:15,210 --> 00:00:22,230 do is cranial model on a computer or on a computer with a GP or in the cloud. 6 00:00:22,230 --> 00:00:28,710 We use a way better devices mostly for running in frames not for training and in order to learn how 7 00:00:28,710 --> 00:00:31,290 to train your own models using your own data. 8 00:00:32,100 --> 00:00:37,980 I have to show you how to do it in Python so that is why we are dealing with Python now. 9 00:00:38,340 --> 00:00:44,620 So we're going to see how to sort of do it our cuts or no cut neural network in Python. 10 00:00:44,940 --> 00:00:51,530 And then later on we shall see how to build other neural networks using pure C quote as well. 11 00:00:51,720 --> 00:00:59,220 But practically you would be using Python to build your neural network cause there is um there's so 12 00:00:59,220 --> 00:01:05,730 many libraries that make it very very easy doing this type of work put a really complex neural networks 13 00:01:05,730 --> 00:01:06,380 with Python. 14 00:01:06,720 --> 00:01:12,930 So there'll be no reason to reinvent what exists in C language but the reason you might want to know 15 00:01:12,930 --> 00:01:18,510 it from C languages as an embedded developer you think in terms of C language. 16 00:01:18,510 --> 00:01:25,800 So I would use C language to explain things to you to show you how things work in C but when you are 17 00:01:25,800 --> 00:01:32,670 in the field doing this work you would most probably use python to train your neural network and run 18 00:01:32,750 --> 00:01:36,390 in France using C language on your embedded device. 19 00:01:36,390 --> 00:01:38,120 So let's get started. 20 00:01:38,340 --> 00:01:44,130 And also I would mention those of you who are already familiar with Python or familiar with building 21 00:01:44,140 --> 00:01:49,140 neural networks with Python you can completely skip this but I should mention though that neural networks 22 00:01:49,140 --> 00:01:52,530 were going to build in these lessons to no use in a library. 23 00:01:52,530 --> 00:01:59,430 We're not going to use cameras or anything if anything just name pi is what we use we use we go according 24 00:01:59,430 --> 00:02:00,890 to what we saw in the theory. 25 00:02:00,890 --> 00:02:06,750 The core lesson and in the theoretical lesson when I was drawing the pseudo code I didn't show any libraries. 26 00:02:07,050 --> 00:02:11,150 So we're going to execute those pure lines here. 27 00:02:11,190 --> 00:02:15,830 We wouldn't use libraries later on we shall see how to use carers intensive flow. 28 00:02:16,170 --> 00:02:20,250 But for now for understanding sake we need to do it raw. 29 00:02:20,990 --> 00:02:21,520 Okay. 30 00:02:21,600 --> 00:02:26,390 I'm going to create a new folder here. 31 00:02:26,650 --> 00:02:28,070 Call this eleven 32 00:02:31,030 --> 00:02:34,270 cut or not. 33 00:02:34,480 --> 00:02:37,690 And in here I'm going to create two food as 34 00:02:41,020 --> 00:02:43,630 accord is data set. 35 00:02:43,630 --> 00:02:52,410 I stole my data sets yet and then images or test images we can store it in this photo and I'm going 36 00:02:52,410 --> 00:02:54,270 to upload that data set. 37 00:02:55,200 --> 00:02:59,430 I'm going to upload a data set to the video so I can download it. 38 00:02:59,430 --> 00:03:05,400 The dataset we're going to use for these lessons um these initial lessons are in a format known as the 39 00:03:05,430 --> 00:03:06,640 dot H5. 40 00:03:07,080 --> 00:03:15,550 So these that the DOT H5 is this is image data that has been converted to dot H5. 41 00:03:15,960 --> 00:03:17,140 Right. 42 00:03:17,190 --> 00:03:19,160 So I'm going to put two of them here. 43 00:03:19,170 --> 00:03:22,160 I have test AMG data at age 5. 44 00:03:22,160 --> 00:03:24,200 Train I am t data dot age 5. 45 00:03:24,300 --> 00:03:28,240 You can think of these as image data so many image data. 46 00:03:28,380 --> 00:03:29,510 So many image. 47 00:03:29,520 --> 00:03:30,020 Um yeah. 48 00:03:30,060 --> 00:03:30,730 Image Data. 49 00:03:30,750 --> 00:03:37,860 Contain any cuts and no cuts converted to this cryptic format that I'm sure some of you are not familiar 50 00:03:37,860 --> 00:03:39,180 with that age 5. 51 00:03:39,300 --> 00:03:42,980 So our images are here compressed and converted. 52 00:03:43,230 --> 00:03:47,790 Later on we shall see how to take images. 53 00:03:47,790 --> 00:03:49,680 Later on we would have. 54 00:03:49,680 --> 00:03:57,990 We would go to cargo and download who download of a maybe sixty thousand images and then make them smaller 55 00:03:58,530 --> 00:04:00,790 and then Crane with those images. 56 00:04:00,990 --> 00:04:04,740 But is so much work and doubt divert us from the neural network. 57 00:04:04,740 --> 00:04:11,220 That's why we started with this already converted image data so later on we're going to download images 58 00:04:11,640 --> 00:04:16,200 and then pretty process the images on our own and then pass them through the neural network. 59 00:04:16,320 --> 00:04:18,020 But for now let's use this. 60 00:04:18,150 --> 00:04:23,280 So we have test AMG data that H5 train AMG data on H5. 61 00:04:23,550 --> 00:04:25,040 Right. 62 00:04:25,060 --> 00:04:26,000 And then um. 63 00:04:26,170 --> 00:04:28,440 So um yeah. 64 00:04:28,650 --> 00:04:34,110 So are we going to have this and when we have been building a new network we'll put our images here 65 00:04:34,110 --> 00:04:41,580 to test a neural network with and before we start to build in the actual neural network we have to um 66 00:04:41,820 --> 00:04:48,570 we have to create some helper functions to help us do things like load the image and maybe compute other 67 00:04:48,570 --> 00:04:49,140 things. 68 00:04:49,140 --> 00:04:55,650 So I'm going to create a python as a python file to hold the helper functions and I'm going to create 69 00:04:56,130 --> 00:04:59,280 another python file to have our actual neural network.