Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated: 1 00:00:03,220 --> 00:00:09,180 Considered these greyscale image. Now this image is made of pixels. 2 00:00:09,270 --> 00:00:16,500 This particular one has three by five so it has 15 pixels so it's quite a small image. 3 00:00:16,500 --> 00:00:24,330 Now each pixel has a value and that's what defines the intensity of the gray color for each pixel. 4 00:00:24,330 --> 00:00:33,030 So in reality these are numbers but our computer display shows them in a color format which is easy 5 00:00:33,030 --> 00:00:35,120 readable by us humans. 6 00:00:35,130 --> 00:00:42,900 Where I'm trying to get is that programs use numbers to store images then the computer display or so 7 00:00:42,900 --> 00:00:46,890 the screen converts these numbers to colors. 8 00:00:46,960 --> 00:00:52,150 Python can also do image processing just like Photoshop does. 9 00:00:52,230 --> 00:00:58,740 Probably it cannot do all the cool stuff that you can do with Photoshop but you can make use of Python 10 00:00:58,740 --> 00:01:06,220 to automate things for instance we'll be using the image processing capabilities of Python to detect faces 11 00:01:06,270 --> 00:01:12,640 from photos, from images and also detect moving objects in videos as well. 12 00:01:12,840 --> 00:01:15,040 So videos are made of images. 13 00:01:15,180 --> 00:01:16,720 So it's the same thing basically. 14 00:01:17,040 --> 00:01:22,600 And Python stores and reads images using arrays of numbers. 15 00:01:22,800 --> 00:01:31,800 For instance this image could be represented as you know a list of three other lists. 16 00:01:34,440 --> 00:01:42,180 So three lists because we have 3 rows in there with pixels and than in each of these you would have like 17 00:01:42,480 --> 00:01:50,950 five numbers and so on and so one and the fifth and the same for the other two lists. 18 00:01:51,060 --> 00:01:57,320 We've got five numbers because we have five columns so five pixels for each row. 19 00:01:57,570 --> 00:01:59,730 That's an image for Python. 20 00:01:59,820 --> 00:02:03,840 And here is where Numpy comes in handy. 21 00:02:03,840 --> 00:02:12,980 So while you can represent images with lists as we did here this is not very efficient because 22 00:02:13,110 --> 00:02:14,150 for big images 23 00:02:14,150 --> 00:02:19,590 lists occupy a lot of memory and therefore they slow down operations on them. 24 00:02:19,690 --> 00:02:27,000 So this is solved by Numpy which is a library, a Python library that provides a multi dimensional 25 00:02:27,000 --> 00:02:28,560 array object. 26 00:02:28,590 --> 00:02:33,980 So let me go ahead and create this array object. 27 00:02:34,320 --> 00:02:37,950 First of all you what you need to do is you need to import Numpy. 28 00:02:38,350 --> 00:02:44,610 And if you have installed Pandas, Numpy should have been installed with Pandas because Pandas is based 29 00:02:44,610 --> 00:02:46,130 on Numpy. 30 00:02:46,770 --> 00:02:51,550 If you haven't installed Numpy just yet just yet, just go had and pip install Numpy. 31 00:02:51,760 --> 00:02:55,990 Ad if for some reason you have some problems on Windows then just go 32 00:02:56,220 --> 00:02:57,570 ahead as I've showed you 33 00:02:57,660 --> 00:02:59,690 and find the precompiled Python libraries. 34 00:02:59,760 --> 00:03:07,720 Go to this site and search for Numpy and then figure out if you are on a 3.5 version of Python then 35 00:03:07,720 --> 00:03:14,320 just get that version and you pointed to this file with pip install and the name of the file. 36 00:03:14,340 --> 00:03:25,020 Great. Now I have Numpy installed so I'll create this multi dimensional object and store it in the N variable. 37 00:03:25,060 --> 00:03:25,750 38 00:03:26,070 --> 00:03:33,020 Let it be numpy.arange and let's say 27. 39 00:03:33,050 --> 00:03:34,820 Execute that. 40 00:03:34,910 --> 00:03:39,390 And print out N. 41 00:03:39,480 --> 00:03:41,650 So this is a Numpy array. That's how it is called. 42 00:03:41,670 --> 00:03:49,370 This particular one is not exactly a multi dimensional array because it only has one dimension so it's 43 00:03:49,580 --> 00:03:50,340 a plane� 44 00:03:50,340 --> 00:03:51,280 It's like a plane list. 45 00:03:51,380 --> 00:03:52,560 A Python list. 46 00:03:52,590 --> 00:03:59,680 But still it's not exactly a list. 47 00:03:59,760 --> 00:04:00,600 Type. Sorry. 48 00:04:01,140 --> 00:04:04,970 So it's a Numpy N dimensional array. 49 00:04:05,130 --> 00:04:09,380 It can have one dimensional two or three. 50 00:04:09,390 --> 00:04:16,370 So we have one, two, three and I'll show all these scenarios. So that was the one dimensional array. 51 00:04:16,380 --> 00:04:23,330 And if you want to print it in a nice form that'd be the array. 52 00:04:23,770 --> 00:04:27,220 Now I'm just creating 53 00:04:27,420 --> 00:04:35,720 a Numpy array using numbers on the fly here but normally you'd have to create arrays from images. 54 00:04:35,880 --> 00:04:38,240 I will do that in just a bit so for now 55 00:04:38,250 --> 00:04:40,610 let's create some arrays manually. 56 00:04:40,790 --> 00:04:47,420 Now let's see what two dimensional arrays� Reshape three by nine. 57 00:04:47,430 --> 00:04:53,550 So we already have this one dimensional array and we want to convert it to a two dimensional array. 58 00:04:53,910 --> 00:04:56,750 If you execute that you get a two dimensional array. 59 00:04:56,850 --> 00:05:00,350 So that's a two dimensional array because it has two dimensions. 60 00:05:00,360 --> 00:05:04,690 So think of it as this image file. 61 00:05:04,800 --> 00:05:08,930 We have two dimensions vertical and horizontal. 62 00:05:08,940 --> 00:05:15,450 Now what is a three dimensional array? Even though a three dimensional arrays are less frequently used 63 00:05:15,450 --> 00:05:15,680 64 00:05:15,830 --> 00:05:17,780 is still good to know them. 65 00:05:17,970 --> 00:05:20,540 And yeah, let me create a three dimensional array. 66 00:05:20,610 --> 00:05:27,810 I could say three by three by three because you know you have 27 elements so three by three by three 67 00:05:27,990 --> 00:05:31,220 gives you twenty seven and yup. 68 00:05:31,280 --> 00:05:33,250 That's a three dimensional rate. 69 00:05:33,360 --> 00:05:36,110 Think of that as a cube that has three dimensions. 70 00:05:36,160 --> 00:05:40,050 So three by three with me and practically you'll see that in just a bit 71 00:05:40,050 --> 00:05:44,880 in this lecture where do we deal with three dimensional array. 72 00:05:45,390 --> 00:05:54,170 So bear with me and you are you able to see the similarities between a Numpy array and a plain Python list 73 00:05:54,170 --> 00:05:55,080 of lists. 74 00:05:55,230 --> 00:06:03,600 So this here would be like a two dimensional array if you like to call it like that. The structure 75 00:06:03,690 --> 00:06:10,800 in the lower level is different between Python lists and Numpy arrays and also Numpy arrays allow you to make 76 00:06:10,810 --> 00:06:16,690 some more efficient operations such as iteration between the array items and so on. 77 00:06:16,800 --> 00:06:25,490 And you can also create a Numpy array out of Python lists. For instance I'll get this list here and I'll create 78 00:06:26,490 --> 00:06:33,930 a new object and then point to Numpy and to covert a list to a Numpy array 79 00:06:34,200 --> 00:06:42,360 you'd want to use Asarray method and then between the brackets goes the object that you want to convert. 80 00:06:43,180 --> 00:06:44,230 81 00:06:44,910 --> 00:06:52,520 And that's an array which is almost exactly like this on in here. 82 00:06:53,070 --> 00:06:56,170 And if you print that you wouldn't be able to see the difference. 83 00:06:56,910 --> 00:07:01,430 You know they look exactly the same but they are not. 84 00:07:01,430 --> 00:07:05,170 Because this is the least of this is a Numpy array. 85 00:07:05,430 --> 00:07:07,050 Great. 86 00:07:07,210 --> 00:07:08,440 Let's move on. 87 00:07:08,880 --> 00:07:09,810 So that's Numpy. 88 00:07:09,810 --> 00:07:14,710 And as I mentioned earlier Numpy is a base library for all the libraries. 89 00:07:15,030 --> 00:07:20,460 Such as Pandas and also OpenCV which is an image processing library. 90 00:07:20,460 --> 00:07:26,910 So Pandas data frames are based on Numpy arrays and OpenCV objects are based on Numpy arrays. 91 00:07:26,910 --> 00:07:27,170 92 00:07:27,210 --> 00:07:34,560 So Pandas, what Pandas does it just adds some cool features in there such as it adds, 93 00:07:34,600 --> 00:07:42,700 it gets capabilities for having table headers and indexes which you can't have in Numpy because Numpy 94 00:07:42,720 --> 00:07:51,150 is meant to be more simple and in a more low level of storing objects and doing operations. 95 00:07:51,240 --> 00:07:54,220 So Numpy is a requirement for many libraries. 96 00:07:54,390 --> 00:07:59,610 And yeah, let's stop this lecture in here and in the next lecture we'll go straight ahead and create 97 00:07:59,670 --> 00:08:03,770 a Numpy array out of our image should be in here. 98 00:08:04,230 --> 00:08:06,640 So this image. Ok, see with there. 9970