1
00:00:03,630 --> 00:00:09,560
Hi, guys, and welcome to the course, this course is called modern computer vision, and they named

2
00:00:09,570 --> 00:00:14,850
it modern computer vision because it is in fact the most modern up to the computer vision, of course,

3
00:00:15,150 --> 00:00:16,530
online right now.

4
00:00:17,130 --> 00:00:23,820
It's a very comprehensive course because it encapsulates all of the open TV, all of the cool, open

5
00:00:23,820 --> 00:00:29,160
TV, classical computer vision theory, as well as all the deep learning modern day terry.

6
00:00:29,580 --> 00:00:31,040
And that's a wide topic.

7
00:00:31,050 --> 00:00:32,970
As you can see, this is a very big course.

8
00:00:32,970 --> 00:00:37,170
It's twenty seven hours, in fact, and I may add content to one little one.

9
00:00:37,680 --> 00:00:39,540
So now let's get started.

10
00:00:40,140 --> 00:00:43,380
So firstly, what exactly is computer vision?

11
00:00:43,860 --> 00:00:50,760
Well, it's an interdisciplinary feel that aims to enable computers or software to gain an understanding

12
00:00:50,760 --> 00:00:53,910
of what is being seen in images and videos.

13
00:00:54,420 --> 00:00:56,610
So many of you may have seen this movie.

14
00:00:56,610 --> 00:00:57,960
It's one of my favorite movies.

15
00:00:58,350 --> 00:01:03,270
Terminator two Judgment Day where Arnold is a robot, the T100 robot.

16
00:01:03,270 --> 00:01:09,660
I believe what too many do, I should say, and this is basically ahead of what he's seeing.

17
00:01:10,020 --> 00:01:14,070
And it was the first computer vision idea I've ever had.

18
00:01:14,280 --> 00:01:19,260
I was a kid when I watched this movie and I imagined thinking, Oh, you can just have cameras and the

19
00:01:19,260 --> 00:01:20,760
camera can make a robot, see.

20
00:01:20,790 --> 00:01:23,610
However, a robot never understands what it actually sees.

21
00:01:23,940 --> 00:01:27,750
That's where all this complex, deep learning software models come into play.

22
00:01:28,350 --> 00:01:30,630
So let's take a look at something else.

23
00:01:30,660 --> 00:01:33,540
Is computer vision artificial intelligence?

24
00:01:34,140 --> 00:01:41,550
Well, yes, it is a big subset of artificial intelligence, which encapsulates a lot of different fields,

25
00:01:41,550 --> 00:01:47,460
including robotics machine learning, which is also data science in a way than deep learning.

26
00:01:47,460 --> 00:01:51,000
Convolutional neural networks are a big part of computer vision.

27
00:01:51,300 --> 00:01:54,260
All of this will make sense to you later on in the course.

28
00:01:54,330 --> 00:01:56,100
Don't worry about these big words right now.

29
00:01:56,130 --> 00:01:57,330
Hope I don't confuse you.

30
00:01:58,140 --> 00:02:01,770
So let's take a look at exactly what computer vision is.

31
00:02:02,190 --> 00:02:05,070
It's an amalgamation of many different fields you can see.

32
00:02:05,070 --> 00:02:12,060
It consists of augmented reality, mathematics and physics, electrical engineering, artificial intelligence,

33
00:02:12,060 --> 00:02:17,550
machine learning, cognitive science, image processing, computer graphics, as well as many other

34
00:02:17,550 --> 00:02:18,390
little fields too.

35
00:02:19,080 --> 00:02:21,210
So hopefully that helps you.

36
00:02:21,240 --> 00:02:26,130
However, you would have been exposed to computer vision applications right now, and you may not have

37
00:02:26,130 --> 00:02:26,850
even known it.

38
00:02:27,150 --> 00:02:30,870
So let's take a look at what some of the things computer vision can do.

39
00:02:31,770 --> 00:02:35,730
So you may be familiar with all these snapshots of Instagram filters.

40
00:02:35,760 --> 00:02:40,950
That is a form of computer vision because they're actually using your input image and then running a

41
00:02:40,950 --> 00:02:41,730
model over it.

42
00:02:41,730 --> 00:02:47,100
To transform the image is optical character recognition, which many of you would have seen if you've

43
00:02:47,100 --> 00:02:52,170
scanned documents and then he automatically can recognize the text in those documents.

44
00:02:53,050 --> 00:02:57,420
License plate reading, which is another form of OCR self-driving cars.

45
00:02:57,420 --> 00:02:59,770
Well, you may not have actually seen these first hand just yet.

46
00:02:59,770 --> 00:03:06,150
They're not on the market PC, although Tesla does have some that does like reverse parking and bunch

47
00:03:06,150 --> 00:03:08,130
of a bunch of highway cool stuff.

48
00:03:08,970 --> 00:03:11,670
So there's also a sports analysis.

49
00:03:11,670 --> 00:03:16,470
You probably would have seen some of these things that cricket is a hockey prediction here, but it

50
00:03:16,470 --> 00:03:20,820
tracks the ball and tells you where it's going to end up for VW decisions.

51
00:03:21,240 --> 00:03:25,950
Then there's a lot of things where you can map players to some sort of geo referencing and have all

52
00:03:25,950 --> 00:03:31,020
of the area that player has covered on the field and a top down view, as well as things like facial

53
00:03:31,020 --> 00:03:32,630
recognition to unlock your phone.

54
00:03:32,680 --> 00:03:35,370
He would have actually seen and used up a lot of things.

55
00:03:36,210 --> 00:03:37,800
And there's actually so much more.

56
00:03:37,830 --> 00:03:42,310
They have things like image recognition, object detection, segmentation API.

57
00:03:42,330 --> 00:03:43,650
These are all examples of it.

58
00:03:43,650 --> 00:03:50,520
Here you can see a lot combining this artistic style of this image onto my image and getting this cool

59
00:03:50,520 --> 00:03:51,330
product here.

60
00:03:51,810 --> 00:03:56,460
Segmentation here, classification, localization and object detection there.

61
00:03:57,420 --> 00:03:59,430
Then you can do things like image similarity.

62
00:03:59,970 --> 00:04:04,140
If you have a big bunch of images and you want to find groups that are clusters that are similarly,

63
00:04:04,140 --> 00:04:08,670
you can use computer vision algorithms for that deepfakes as well.

64
00:04:08,700 --> 00:04:13,320
Here's an example of Nicolas Cage's fierce being overlaid onto his.

65
00:04:13,710 --> 00:04:16,230
And you can see it's a bit creepy and it's a bit dangerous.

66
00:04:16,230 --> 00:04:19,050
Deepfakes are a very hot topic.

67
00:04:19,770 --> 00:04:22,290
Now you can look at body pose estimation as well.

68
00:04:22,290 --> 00:04:28,740
We can actually identify where each limb is and the angle it's oriented to, as well as image generation

69
00:04:28,740 --> 00:04:33,090
to generate fake anime characters or many other fake types of images.

70
00:04:33,810 --> 00:04:36,570
So the computer vision applications are endless.

71
00:04:37,020 --> 00:04:43,620
This is a sinister robot for a very cool computer vision company has put up and is basically hundreds

72
00:04:43,620 --> 00:04:44,340
of more of these.

73
00:04:44,730 --> 00:04:47,670
You can imagine a whole wide computer vision applications.

74
00:04:48,840 --> 00:04:50,400
So why should you do this course?

75
00:04:50,790 --> 00:04:52,350
And what exactly are you going to learn?

76
00:04:52,860 --> 00:04:56,940
Well, this course is separated into two sections.

77
00:04:57,310 --> 00:04:59,250
There's the classical computer vision or.

78
00:04:59,390 --> 00:05:05,270
In KBE section, where we took an in-depth look at all of the traditional computer vision algorithms

79
00:05:05,600 --> 00:05:09,200
that have been developed from the 1970s to even modern day.

80
00:05:09,440 --> 00:05:13,610
It's a lot of work still being done, so we cover all of those things.

81
00:05:13,610 --> 00:05:18,530
Do all of those things are separate to deep learning because deep learning has changed the game?

82
00:05:18,980 --> 00:05:26,330
Deep learning has basically allowed us to create very complex, very cool image models that can do so

83
00:05:26,330 --> 00:05:30,290
many different things like object detection that you're seeing here, as well as tracking.

84
00:05:30,590 --> 00:05:32,000
So who should do this course?

85
00:05:32,690 --> 00:05:37,340
Well, anyone who has a strong interest in computer vision can do this course.

86
00:05:37,760 --> 00:05:43,130
However, the main target I think for this course would be college students, and this can take any

87
00:05:43,130 --> 00:05:47,180
level basically be a CMC or Ph.D., all of those guys.

88
00:05:47,510 --> 00:05:53,390
If you have a computer vision course or a computer vision project, there's a lot of resources out there

89
00:05:53,390 --> 00:05:56,570
that will confuse you when you're getting started in computer vision.

90
00:05:57,110 --> 00:06:02,150
That's why I angled this course, basically starting at classical computer vision and going onto deep

91
00:06:02,150 --> 00:06:07,370
learning that builds a foundation and gives you a very good, proper knowledge of computer vision.

92
00:06:08,270 --> 00:06:11,110
High school students and hobbyists can very well take this course.

93
00:06:11,120 --> 00:06:16,340
You can get a lot out of it and build simple prototypes using the knowledge you gain in discourse.

94
00:06:17,180 --> 00:06:23,870
And anyone with a computer or software engineering background who wants to get started in computer vision,

95
00:06:23,870 --> 00:06:27,750
like data scientists or software engineers, they can do this course as well.

96
00:06:27,770 --> 00:06:32,150
I think that's a big chunk of people who might want to be doing this course as well.

97
00:06:32,600 --> 00:06:34,130
So about me?

98
00:06:35,090 --> 00:06:40,460
OK, well, firstly, I was an electrical and computer engineer for seven years as a radio frequency

99
00:06:40,790 --> 00:06:43,430
engineer doing cell site planning and optimization.

100
00:06:43,850 --> 00:06:47,830
Then I did my masters in artificial intelligence at the University of Edinburgh.

101
00:06:47,840 --> 00:06:54,230
That was in 2014, then working in computer vision for the last six, almost seven years now.

102
00:06:54,650 --> 00:06:59,330
I've looked at several London startups and even co-founded my own company at one time.

103
00:06:59,690 --> 00:07:02,030
However, we went bust, so that's OK.

104
00:07:02,090 --> 00:07:03,380
It's a good learning experience.

105
00:07:04,460 --> 00:07:10,340
I've created several courses Udemy courses on computer vision as well as on some other platforms as

106
00:07:10,340 --> 00:07:10,730
well.

107
00:07:11,060 --> 00:07:16,700
And no, I'm currently a senior computer vision engineer at a company called Davos.

108
00:07:17,840 --> 00:07:19,520
So these are my Udemy courses.

109
00:07:20,180 --> 00:07:23,990
I know you can see I have a decent instructor rating of 4.3.

110
00:07:24,350 --> 00:07:29,810
It should be higher, but the reason it isn't higher, it's mainly because these two courses my big

111
00:07:29,930 --> 00:07:30,950
computer vision courses.

112
00:07:30,950 --> 00:07:33,170
Here, you can see how many reviews I've got on this.

113
00:07:33,770 --> 00:07:34,820
They're a bit outdated.

114
00:07:35,000 --> 00:07:37,180
This one was created in 2016.

115
00:07:37,610 --> 00:07:41,780
It was basically an open CV course, and I updated it later on to OpenSea before.

116
00:07:42,170 --> 00:07:45,700
However, I do cover a fair amount of open CV.

117
00:07:45,710 --> 00:07:49,100
However, this course covers even more topics than that course and open CV.

118
00:07:49,700 --> 00:07:55,640
And then this course was created this deep learning computer vision course in twenty eighteen, and

119
00:07:55,640 --> 00:07:56,930
that's about four years ago now.

120
00:07:57,350 --> 00:08:02,300
And even though it's still a rather useful course, I do reference it a lot in my project something

121
00:08:02,300 --> 00:08:03,200
from time to time.

122
00:08:04,070 --> 00:08:10,880
It's a bit outdated, and it doesn't have a lot of the modern object detection theory and projects in

123
00:08:10,880 --> 00:08:11,120
it.

124
00:08:11,600 --> 00:08:17,570
So basically redone updated that course and that course was only carries TensorFlow.

125
00:08:17,870 --> 00:08:22,940
This course now uses PyTorch along with TensorFlow and Keras, so that's a big plus.

126
00:08:23,390 --> 00:08:29,600
So basically, I'm just going to sum up why I made this course amid Middle-schoolers to basically update

127
00:08:29,600 --> 00:08:35,930
those to all the Udemy computer vision courses and have combined both of them together, as well as,

128
00:08:35,930 --> 00:08:41,210
I think a lot of the student feedback and a lot of the feedback was basically broken, outdated, cold

129
00:08:41,480 --> 00:08:45,710
and hard, difficult to set up libraries and installations.

130
00:08:46,130 --> 00:08:52,430
So to solve that, everything now is being hosted on a run on Google collab, so there's no messy installs

131
00:08:52,430 --> 00:08:53,600
of virtual machine setups.

132
00:08:54,290 --> 00:08:57,200
All the code is up to date as of 2022.

133
00:08:57,650 --> 00:09:01,760
Things do break from time to time when PyTorch or TensorFlow of the their visions.

134
00:09:02,150 --> 00:09:05,180
However, it's relatively easy to fix for me.

135
00:09:05,180 --> 00:09:10,490
Now I can just go back and see what's what change in the library, make those updates and I'll keep

136
00:09:10,490 --> 00:09:16,160
the code up to date for all of you guys and I cover all the key areas, at least what I consider all

137
00:09:16,160 --> 00:09:19,860
the key areas in computer vision for modern day applications.

138
00:09:20,300 --> 00:09:23,780
And as I said, it's what TensorFlow and PyTorch.

139
00:09:24,590 --> 00:09:31,130
And so basically it's a big course with open CV and deep learning modules in one single 27 hour course.

140
00:09:31,490 --> 00:09:34,340
So what are the requirements for doing discourse?

141
00:09:34,490 --> 00:09:40,310
Well, you will need an internet connection just because to Google Cloud is a cloud platform.

142
00:09:40,400 --> 00:09:42,200
You don't have to have a fast internet connection.

143
00:09:42,200 --> 00:09:45,290
Any broadband connection would work even.

144
00:09:45,470 --> 00:09:48,500
I mean, I see at least two megabit, but probably even less would.

145
00:09:48,500 --> 00:09:53,120
Work is just going to be a bit slower when you're downloading or loading different things.

146
00:09:54,560 --> 00:09:56,240
You don't need much, Matt.

147
00:09:56,360 --> 00:09:58,170
However, it would help.

148
00:09:58,190 --> 00:09:59,120
So some high school.

149
00:09:59,410 --> 00:10:05,620
That would be beneficial, as well as programming, you don't necessarily need to understand Python

150
00:10:05,620 --> 00:10:08,170
or programming in general to do this course.

151
00:10:08,530 --> 00:10:11,080
All of the code is explained and highly commented.

152
00:10:11,470 --> 00:10:13,720
So it's kind of self-explanatory.

153
00:10:14,020 --> 00:10:18,090
I know a lot of the code later on, and deep learning part is a bit complicated.

154
00:10:18,100 --> 00:10:20,260
However, I do explain it line by line.

155
00:10:20,710 --> 00:10:22,870
So hopefully that should help.

156
00:10:23,410 --> 00:10:29,860
And generally, I want you to have enthusiasm for it and computer vision because enthusiasm is what

157
00:10:29,860 --> 00:10:31,600
will motivate you to do this course.

158
00:10:31,960 --> 00:10:37,150
It's a long course, however, it's a very comprehensive course and you will get a lot out of it.

159
00:10:37,570 --> 00:10:42,340
So this is just an overview of all the topics I covered in the open CV section.

160
00:10:42,670 --> 00:10:46,660
I'm not going to read out all of these because it's just, I mean, you can read it yourself, so you

161
00:10:46,660 --> 00:10:50,410
can just pause the video if you want to take a closer look of all of them.

162
00:10:51,100 --> 00:10:54,280
And these are the deep learning topics as well that we cover.

163
00:10:54,640 --> 00:10:59,950
However, each of these has like sometimes 10 to 20 sub lectures involved.

164
00:11:00,340 --> 00:11:01,690
Some of them are much less so.

165
00:11:02,620 --> 00:11:09,610
So it's basically covering everything you need to know for understanding deep learning, using the word

166
00:11:09,610 --> 00:11:11,680
PyTorch and TensorFlow with Garrus.

167
00:11:12,460 --> 00:11:16,610
So in the next section, I'll just take a deeper look at the course overview.

168
00:11:16,630 --> 00:11:21,910
Again, I'm not going to read out every slide, every lecture title, but I'm going to go into the topics

169
00:11:21,910 --> 00:11:22,420
a bit more.

170
00:11:22,900 --> 00:11:25,480
So thank you for watching and I'll see you in that section.