1
00:00:00,300 --> 00:00:01,530
Hi, welcome back.

2
00:00:01,620 --> 00:00:08,400
So firstly, the reason I split this lesson into two videos is because what I'm going to showcase next

3
00:00:08,430 --> 00:00:14,580
is the body pose estimation models that belong to the that that run to the library.

4
00:00:15,060 --> 00:00:19,440
And I didn't want to mix it too much with the mass, you know, because it's a separate topic.

5
00:00:19,440 --> 00:00:26,310
So I'm going to create a separate video entitled so that if you're searching for body pause in discourse,

6
00:00:26,310 --> 00:00:29,550
you can easily find it in its own separate video here.

7
00:00:30,180 --> 00:00:35,490
So, yes, you can see this is the image of me looking fit test image of a footballer.

8
00:00:36,400 --> 00:00:40,410
Next, we can follow the same patterns that we did before where we lowered.

9
00:00:40,920 --> 00:00:47,340
Basically, this is the config file, and then we set it to visualize and create the default predictor

10
00:00:47,340 --> 00:00:48,120
as well.

11
00:00:48,720 --> 00:00:55,200
And we can run this here, and you can see we can quickly identify that this is a person here.

12
00:00:56,220 --> 00:00:57,700
You can see this is a very cool.

13
00:00:57,700 --> 00:01:02,460
You can see each limb is in a different color, so you can see this is a very cool way.

14
00:01:02,820 --> 00:01:06,040
This is bodies pause and the angle is actually correct as well.

15
00:01:06,060 --> 00:01:13,280
Similarly, here between the shoulders, the eyes, fierce nose arms is probably a person in the background

16
00:01:13,290 --> 00:01:15,570
doing something I actually don't recall.

17
00:01:15,630 --> 00:01:16,240
Yeah, there is.

18
00:01:16,400 --> 00:01:19,560
There's a person right there in the background and actually who's quite leery.

19
00:01:20,130 --> 00:01:23,760
And while it got the body, police estimate quite wrong.

20
00:01:23,880 --> 00:01:26,250
But it is quite cool that it actually tried it anyway.

21
00:01:27,150 --> 00:01:30,780
So you can see this is how we inference on a body pose estimation.

22
00:01:31,290 --> 00:01:40,230
Now back to basically panopticon segmentation, which basically does every object different.

23
00:01:40,230 --> 00:01:43,060
A different color effectively can.

24
00:01:43,260 --> 00:01:47,190
You can see that you have separate persons, each person is in a different color as well.

25
00:01:47,610 --> 00:01:50,310
It is a bit like instance segmentation.

26
00:01:50,310 --> 00:01:57,120
However, it goes into more granular detail and you can see how much detail it tries to predict here.

27
00:01:57,600 --> 00:01:59,370
Playing field right sportsbook.

28
00:01:59,760 --> 00:02:00,260
Pretty cool.

29
00:02:00,450 --> 00:02:02,550
It should be football, but I mean, it's possible.

30
00:02:02,590 --> 00:02:03,270
It's fair enough.

31
00:02:03,820 --> 00:02:10,200
It's probably a generic category of basketballs and footballs and tennis balls, cricket balls.

32
00:02:10,200 --> 00:02:13,500
Perhaps, so you can see that looks quite well.

33
00:02:14,460 --> 00:02:18,150
Now what I actually can do, although I didn't have it, don't have a dunk.

34
00:02:18,160 --> 00:02:20,520
It takes quite some time to generate.

35
00:02:20,940 --> 00:02:26,790
Is that you can actually run a pin up tick segmentation video so you can load this video here and you

36
00:02:26,790 --> 00:02:27,690
can just cut.

37
00:02:27,870 --> 00:02:31,350
Is an f ffmpeg cut the first six seconds of it here.

38
00:02:31,350 --> 00:02:33,900
Actually, let's try it and see if it works.

39
00:02:35,580 --> 00:02:38,520
It just takes a bit of time to do.

40
00:02:42,450 --> 00:02:43,830
But let's see.

41
00:02:44,820 --> 00:02:45,480
Oh, it's the.

42
00:02:48,930 --> 00:02:54,030
I guess this YouTube Library is broken, so I'll probably fix that for you guys.

43
00:02:54,720 --> 00:02:59,730
And you can hopefully on a notebook, it works well, but that's it for this lesson.

44
00:03:00,480 --> 00:03:01,380
Thank you for watching.

45
00:03:01,390 --> 00:03:08,520
And in the next section, we'll take a look at creating a mask are seen in a different mask, our CNN

46
00:03:08,910 --> 00:03:11,130
to recognize various ships.

47
00:03:11,310 --> 00:03:12,810
So stay tuned for that lesson.

48
00:03:12,930 --> 00:03:13,770
Thank you for watching.

49
00:03:13,950 --> 00:03:14,220
But.