1
00:00:01,240 --> 00:00:08,194
To create your AI activity timer, you 
will train a machine learning, or ML, model

2
00:00:08,194 --> 00:00:13,080
to recognise when you’re doing 
different movements or activities. 

3
00:00:13,080 --> 00:00:18,960
You’ll then combine that model with some 
ready-made code for an activity timer,  

4
00:00:18,960 --> 00:00:26,800
before downloading it to your 
micro:bit and using it in real life.

5
00:00:26,800 --> 00:00:33,960
Click ‘Open in micro:bit CreateAI’ to launch the project.

6
00:00:33,960 --> 00:00:41,120
This project comes with 6 samples of movement 
data for walking, 6 samples of movement data  

7
00:00:41,120 --> 00:00:47,200
for jumping up and down, and 6 samples of 
movement data for staying fairly still.

8
00:00:47,200 --> 00:00:54,200
You will add more samples by 
recording your own movement data.

9
00:00:54,200 --> 00:01:02,703
micro:bit CreateAI collects movement data
samples using the accelerometer,  
or movement sensor, on the micro:bit.

10
00:01:05,726 --> 00:01:09,734
You will wear a micro:bit and battery pack on your wrist or ankle

11
00:01:09,734 --> 00:01:15,120
so that you can move freely
to record your own movement data samples. 

12
00:01:15,120 --> 00:01:20,400
To get started, connect the 
ankle-worn micro:bit to CreateAI. 

13
00:01:20,400 --> 00:01:23,816
We call this the data collection micro:bit.

14
00:01:23,816 --> 00:01:31,640
If your computer has Bluetooth enabled then you will just need 1 micro:bit with a battery pack and a USB data lead.

15
00:01:31,640 --> 00:01:36,640
If you don’t have a Bluetooth connection, 
you’ll be prompted to use 2 micro:bits.

16
00:01:36,640 --> 00:01:41,400
The second micro:bit will remain 
connected to the USB cable and act as  

17
00:01:41,400 --> 00:01:50,600
a radio link to the data collection micro:bit.
Follow the instructions on screen to connect.

18
00:01:50,600 --> 00:01:57,817
Once your micro:bit is connected, you will see the lines on the live graph change as  you move your micro:bit about.

19
00:01:57,817 --> 00:02:02,297
You’re now ready to add your own movement data samples.

20
00:02:02,297 --> 00:02:06,263
As this project already includes some data samples,

21
00:02:06,263 --> 00:02:16,912
we suggest you just add 1 more sample for each action for now, and spend more time collecting and analysing data later. 

22
00:02:16,912 --> 00:02:24,800
Make sure your data collection micro:bit is attached to the inside of the ankle, with button B on top. 

23
00:02:24,800 --> 00:02:29,280
To add data to a specific action, 
select it by clicking on it. 

24
00:02:29,280 --> 00:02:34,440
You will get a 3 second countdown 
before a 1 second recording starts. 

25
00:02:34,440 --> 00:02:40,080
Click record and start moving right away 
to ensure you get a clean data sample. 

26
00:02:40,080 --> 00:02:43,960
A clean sample is one where you 
are moving for the entire sample,  

27
00:02:43,960 --> 00:02:48,593
you don’t start late or finish moving early.

28
00:02:48,593 --> 00:02:55,519
Next try adding an extra data sample to the ‘jumping’ data set and the ‘being still’ data set.

29
00:02:55,519 --> 00:03:05,040
Select them by clicking on the action, then click record and jump or stay quite still as you record the samples. 

30
00:03:05,040 --> 00:03:16,521
You’ll notice on the 'being still' samples 
that the x,y,z lines change places depending on the angle of the attached micro:bit.

31
00:03:16,521 --> 00:03:26,080
We don’t have a lot of data right now, but we do have enough to train our own  machine learning model using CreateAI. 

32
00:03:26,080 --> 00:03:33,720
So click ‘Train model’ to use the 
current data to build an ML model.

33
00:03:33,720 --> 00:03:41,363
The tool now builds a mathematical model that should recognise different actions when you move your micro:bit.

34
00:03:41,363 --> 00:03:45,646
As soon as the model has been trained, you will see the Testing model page.

35
00:03:45,646 --> 00:03:50,721
Now use the data collection micro:bit to test how well the model is working.

36
00:03:50,721 --> 00:03:59,720
It should still be connected to the tool, and you’ll see that as you move it, CreateAI is estimating what action you are doing.

37
00:04:02,360 --> 00:04:09,698
Try out each of the actions to see 
both the estimated action and the certainty bar graph change.

38
00:04:11,257 --> 00:04:19,126
The % on the certainty bar graph shows how confident the model is that you are doing each action. 

39
00:04:21,600 --> 00:04:26,520
You may notice your model is not estimating 
some actions accurately, or maybe it is  

40
00:04:26,520 --> 00:04:32,720
working well for one action but not the other, 
so after exploring how it is currently working,  

41
00:04:32,720 --> 00:04:40,360
it is a good idea to click on ‘Edit 
data samples’ and improve your model. 

42
00:04:40,360 --> 00:04:46,360
Machine learning models usually work best with 
MORE data, so record some extra samples for each  

43
00:04:46,360 --> 00:04:54,960
of the actions, or focus on collecting more data 
for the action that was problematic in testing. 

44
00:04:54,960 --> 00:05:01,935
You can record one sample at a time or 
you can record 10 samples in sequence. 

45
00:05:10,160 --> 00:05:23,081
Clean data samples also help an ML model work 
better so examine your data set and identify any data samples that could confuse the model.

46
00:05:26,003 --> 00:05:29,640
You can delete these by pressing X.

47
00:05:30,840 --> 00:05:39,423
Once you’ve added more data and checked your data set, click ‘Train model’ again to use your amended data set. 

48
00:05:40,483 --> 00:05:45,992
Then test the model again on the ‘Testing model’ page.

49
00:05:48,555 --> 00:05:54,640
Once you’re happy with how the ML model is behaving, you can use it with the ready-made project code. 

50
00:05:54,640 --> 00:06:02,480
Click on ‘Edit in MakeCode’ to see the code 
blocks in a special version of Microsoft MakeCode. 

51
00:06:02,480 --> 00:06:09,560
You can always return to CreateAI using 
the arrow in the top left of the screen. 

52
00:06:09,560 --> 00:06:19,240
These code blocks use the model you 
have created within an exercise timer. 

53
00:06:19,240 --> 00:06:26,160
The code uses 3 variables to keep track 
of how long you've been doing each action. 

54
00:06:26,160 --> 00:06:33,680
When the program first runs it 
sets these timer variables to 0. 

55
00:06:33,680 --> 00:06:40,120
The 'on ML start' blocks are triggered when the ML 
model decides you have started a specific action. 

56
00:06:40,120 --> 00:06:49,250
They show different icons on the micro:bit's LED display depending on the action it has estimated you are doing.

57
00:06:50,289 --> 00:07:01,280
The 'on ML stop' blocks are triggered when the ML model decides you have finished an action, 
in this case walking, jumping or being still. 

58
00:07:01,280 --> 00:07:06,880
Code inside each block clears the screen and 
adds the duration of the action that has just  

59
00:07:06,880 --> 00:07:13,440
finished to the variable storing 
the total times for each action. 

60
00:07:13,440 --> 00:07:19,440
The ML model works with the code to allow you 
to view the total time spent on each action. 

61
00:07:19,440 --> 00:07:23,040
Press button A to see the estimate 
of how long you were walking. 

62
00:07:23,040 --> 00:07:28,160
Press button B to see how long the 
model estimated you were jumping. 

63
00:07:28,160 --> 00:07:34,080
To see the estimated duration you have 
been still press A and B together. 

64
00:07:34,080 --> 00:07:44,111
The timer counts in milliseconds, thousandths of a second, so the number shown is divided by 1000 to show a time in seconds.

65
00:07:44,111 --> 00:07:53,160
To make your AI activity timer run on your micro:bit, you just need to download this code to a micro:bit. 

66
00:07:53,160 --> 00:08:00,642
If you don’t have another micro:bit available,
simply replace the code currently on the data collection micro:bit with the project code.

67
00:08:01,646 --> 00:08:05,000
Now you can test the project out in real life.

68
00:08:05,000 --> 00:08:10,040
Do the correct icons display 
when you are exercising or not? 

69
00:08:10,040 --> 00:08:15,800
You can test if the timer code is working 
well with the model in 3 easy steps:

70
00:08:15,800 --> 00:08:19,160
Press the reset button.
Jump for 30 seconds.

71
00:08:19,160 --> 00:08:25,540
Then press button B.
You should see the number 30 scroll across your display.

72
00:08:25,540 --> 00:08:28,560
You’re now ready to connect to CreateAI,  

73
00:08:28,560 --> 00:08:34,560
collect your own data, use it to train, test and 
improve a machine learning model, and then you  

74
00:08:34,560 --> 00:08:40,600
can combine this model with the ready-made 
code and try it out on your own micro:bit. 

75
00:08:40,600 --> 00:08:48,379
If you’re looking for ways to personalise this even more try adding some different actions like running or dance steps.

76
00:08:48,379 --> 00:08:52,480
Enjoy!