1
00:00:03,000 --> 00:00:08,000
His directorial provides a brief introduction about YOLO V8.

2
00:00:08,000 --> 00:00:14,000
We will first look at the list of contents which we will cover in this video tutorial.

3
00:00:15,000 --> 00:00:21,000
So following are the list of contents or the objectives which we will cover in this video tutorial.

4
00:00:21,000 --> 00:00:28,000
We have divided this video complete lecture into four parts starting from the first, which is what

5
00:00:28,000 --> 00:00:29,000
is YOLO.

6
00:00:29,000 --> 00:00:34,000
The second part, which we will covered in this video tutorial is what is YOLO V8.

7
00:00:34,000 --> 00:00:39,000
In the third part we will discuss what are the key features of YOLO V8.

8
00:00:39,000 --> 00:00:45,000
In the fourth part we will discuss what are the reasons or what are the benefits of using Adobe Air.

9
00:00:46,000 --> 00:00:49,000
So starting from the first, which is what is YOLO?

10
00:00:50,000 --> 00:00:54,000
So YOLO stands for you only look once.

11
00:00:54,000 --> 00:01:02,000
YOLO is a family of computer vision model which was introduced by Joseph Redmon and those the Santosh

12
00:01:02,000 --> 00:01:07,000
Devala, Ross Khurshid and Ali Farhadi introduced.

13
00:01:07,000 --> 00:01:07,000
YOLO.

14
00:01:08,000 --> 00:01:11,000
YOLO is gaining popularity among introduce.

15
00:01:11,000 --> 00:01:14,000
Day by day because it is being.

16
00:01:14,000 --> 00:01:23,000
Since its launch from 2015 till now, eight different versions of YOLO has been introduced.

17
00:01:23,000 --> 00:01:30,000
So the YOLO has been gaining popularity because of its accuracy while maintaining a small model size.

18
00:01:30,000 --> 00:01:37,000
YOLO model can be trained on a single GPU, which makes his assets, which makes it accessible to a

19
00:01:37,000 --> 00:01:39,000
wide range of developers.

20
00:01:39,000 --> 00:01:46,000
Since its launch in 2015 by Joseph Redmon, time by time, YOLO has been upgraded by the computer vision

21
00:01:46,000 --> 00:01:49,000
community from versions 1 to 4.

22
00:01:49,000 --> 00:01:56,000
YOLO was maintained in C code in a custom deep learning framework written by Redmond called Chakra.

23
00:01:57,000 --> 00:01:59,000
In the last two years, YOLO.

24
00:01:59,000 --> 00:02:00,000
YOLO.

25
00:02:00,000 --> 00:02:07,000
Yolo v six Yolo V7 has average around the world out of their own PyTorch based implementation.

26
00:02:07,000 --> 00:02:14,000
Each model has brought new SOTA techniques, new SOTA techniques that continue to push more like model

27
00:02:14,000 --> 00:02:17,000
accuracy and efficiency.

28
00:02:18,000 --> 00:02:24,000
So here we have learned the few reasons that why is YOLO gaining popularity?

29
00:02:24,000 --> 00:02:28,000
One of the reason is that that YOLO has been in a continuous upgrade.

30
00:02:28,000 --> 00:02:34,000
There is a lot of research has been done by the computer vision community members, members on YOLO

31
00:02:34,000 --> 00:02:40,000
and YOLO has been upgraded after every few months like it has been a few months back that Yolov7 was

32
00:02:40,000 --> 00:02:40,000
introduced.

33
00:02:40,000 --> 00:02:42,000
And then now we have YOLO V8.

34
00:02:42,000 --> 00:02:48,000
So after few months we might have the YOLO V9 So there is a continuous research plus.

35
00:02:49,000 --> 00:02:56,000
Because of accuracy and maintaining a small model size YOLO model can be trained on a single GPU, so

36
00:02:56,000 --> 00:02:59,000
that's make it accessible to a wide range of developers around the world.

37
00:02:59,000 --> 00:03:04,000
So that is the other reason that YOLO is being gaining the popularity.

38
00:03:04,000 --> 00:03:07,000
So now we will discuss what is YOLO V8.

39
00:03:07,000 --> 00:03:08,000
So geology.

40
00:03:08,000 --> 00:03:13,000
It is the latest version of YOLO which was released on January 10th, 2023.

41
00:03:14,000 --> 00:03:22,000
So YOLO V8 outperforms all the other versions of YOLO in terms of accuracy, speed and at training time.

42
00:03:22,000 --> 00:03:30,000
So generally in YOLO V8, that mean average precision on which basically we judge a YOLO model mean

43
00:03:30,000 --> 00:03:38,000
average precision of 53.7 has been marked by YOLO V8, which is highest ever in the YOLO history.

44
00:03:38,000 --> 00:03:44,000
So YOLO is the new state of the art model that can be used for object detection, image classification

45
00:03:44,000 --> 00:03:49,000
and also, for instance, segmentation tasks as in YOLO.

46
00:03:49,000 --> 00:03:53,000
Here the issue of prolonged training is what is fixed as well.

47
00:03:53,000 --> 00:04:01,000
Like I have done comparison of YOLO V8 with Yolov7 pose using license plate detection dataset.

48
00:04:01,000 --> 00:04:10,000
And what are my chances are that in comparison to V8 in comparison to Yolov7 YOLO V8 takes less training

49
00:04:10,000 --> 00:04:17,000
time, less YOLO V8 is more accurate than Yolov7, so Yolov7 takes more training time and it is less

50
00:04:17,000 --> 00:04:22,000
accurate while YOLO v8 take less training time and it is more accurate than YOLO.

51
00:04:22,000 --> 00:04:31,000
V7 okay, so the trade off between training time and precision is somewhat achieved in YOLO V8 as well.

52
00:04:32,000 --> 00:04:35,000
So here are the key features of using YOLO V8.

53
00:04:35,000 --> 00:04:40,000
So now let's look at some of the key features of using YOLO V8.

54
00:04:40,000 --> 00:04:44,000
The key feature of YOLO V8 is its access extensibility.

55
00:04:44,000 --> 00:04:50,000
YOLO V8 is basically designed as a framework that support all the previous version of YOLO.

56
00:04:50,000 --> 00:04:55,000
So make it easier to switch between different versions and compare their performance.

57
00:04:55,000 --> 00:04:57,000
So I can switch with YOLO v3.

58
00:04:57,000 --> 00:05:03,000
YOLO V5 and compare the performances of YOLO V8 with Yolov5 and YOLO V3.

59
00:05:03,000 --> 00:05:10,000
YOLO V8 includes a new backbone network which includes a series of convolution layer and also you.

60
00:05:11,000 --> 00:05:18,000
It introduces a new anchor for detection head, which makes the YOLO V8 more accurate than previous

61
00:05:18,000 --> 00:05:19,000
version.

62
00:05:19,000 --> 00:05:25,000
YOLO V8 is highly efficient and can run a variety of platforms on CPU and GPU as well.

63
00:05:25,000 --> 00:05:33,000
And while the V7 can also be run on GPU and VR, it can also be run on CPU and Yolov7 can also be run

64
00:05:33,000 --> 00:05:35,000
on CPU and GPU.

65
00:05:35,000 --> 00:05:42,000
And while we can get good performances or result and fastest training time on GPU as you all know as

66
00:05:42,000 --> 00:05:42,000
well.

67
00:05:42,000 --> 00:05:48,000
So basically anchor free detections make YOLO V8 more accurate than previous version version.

68
00:05:48,000 --> 00:05:54,000
So as the Ultralytics doc says, V8 is a cutting edge state of the art model.

69
00:05:54,000 --> 00:06:01,000
YOLO V8 introduces new features and improved improvements to further boost performance as well as the

70
00:06:01,000 --> 00:06:02,000
reliability.

71
00:06:02,000 --> 00:06:03,000
So YOLO.

72
00:06:03,000 --> 00:06:09,000
It uses anchor free detection and new convolution layers to make predictions more accurate.

73
00:06:09,000 --> 00:06:14,000
So anchor free detection makes your loving aid model more accurate and faster than previous version.

74
00:06:15,000 --> 00:06:19,000
So in YOLO, we aired the issue of prolonged training time is one fixed?

75
00:06:20,000 --> 00:06:25,000
Whereas as I told you previously, the trade off between training time and precision is achieved in

76
00:06:25,000 --> 00:06:26,000
YOLO V8.

77
00:06:26,000 --> 00:06:31,000
I have told you in the previous slide as well, like how we did it basically by considering the license

78
00:06:31,000 --> 00:06:32,000
plate detection dataset.

79
00:06:32,000 --> 00:06:38,000
We check it and New Backbone Network, which includes a series of convolution layers and a new anchor

80
00:06:38,000 --> 00:06:42,000
for detection head and a new loss function make YOLO V8 much faster.

81
00:06:43,000 --> 00:06:46,000
So these are the reasons in YOLO.

82
00:06:46,000 --> 00:06:48,000
YOLO is basically anchor free.

83
00:06:48,000 --> 00:06:53,000
Yolo V here does not predict based on bounding box anchors, which is what the previous models used

84
00:06:53,000 --> 00:06:54,000
to do.

85
00:06:55,000 --> 00:06:59,000
So Coco dataset is the benchmark dataset that people use.

86
00:06:59,000 --> 00:07:02,000
We use for object and object detection on.

87
00:07:02,000 --> 00:07:06,000
So it has been validated on the Coco dataset.

88
00:07:07,000 --> 00:07:11,000
Here are what are the reasons or what are the benefits of using YOLO?

89
00:07:11,000 --> 00:07:12,000
V eight.

90
00:07:12,000 --> 00:07:18,000
So here I listed a few reasons that why I should use YOLO V8 or you should use YOLO V8 for your computer

91
00:07:18,000 --> 00:07:19,000
vision project.

92
00:07:19,000 --> 00:07:25,000
YOLO V8 has a high rate of accuracy measured by Coco and Roboflow for YOLO.

93
00:07:25,000 --> 00:07:31,000
V8 outperforms all other YOLO models in terms of speed and accuracy means it achieves a mean average

94
00:07:31,000 --> 00:07:33,000
precision of 53.7.

95
00:07:33,000 --> 00:07:40,000
Yolo V8 can be installed in two ways from the source, like by coding the GitHub repo or by PIP by writing

96
00:07:40,000 --> 00:07:46,000
PIP install Ultralytics, which is the first iteration of YOLO to have an official package like PIP

97
00:07:46,000 --> 00:07:47,000
install Ultralytics.

98
00:07:47,000 --> 00:07:52,000
So New Backbone Network which includes series of convolution layer, new anchor projection head and

99
00:07:52,000 --> 00:07:59,000
a new loss function make YOLO V8 much faster, reliable and more precise and efficient.

100
00:07:59,000 --> 00:08:05,000
YOLO V8 does not predict based on bonding of sucker anchors like I've already told you by the other

101
00:08:05,000 --> 00:08:06,000
models used to do this.

102
00:08:06,000 --> 00:08:09,000
So this is a brief introduction about YOLO V8.

103
00:08:09,000 --> 00:08:11,000
Let's move towards the next tutorial.