1 of 30

Creating a Data Set

Machine Learning - Session 3

2 of 30

Skills I Have Already…

This is the start of your Machine Learning journey. You will be finding out what it is and how it works.

Lesson 1�Introduction to Machine Learning

Lesson 4�Applying the Data Set

Lesson 3�Creating a Data Set

Lesson 2�Trying out Machine Learning

Machine Learning

Skills I Will Learn Next…

Now you know how Machine Learning works, you can think about how you use it and what you need to remember to use it effectively.

Patterns

Prediction

Machine Learning

Label

Features

Algorithm

Training

My iPad Learner Journey

3 of 30

What did we learn last session?

Teaching our Robot

Machine Learning Vocabulary

Testing our Robot

Creating a Data Set

4 of 30

In this session…

This lesson we will be collecting a data set for the CO-ML app that Apple have given us access to. This will include…

  • How it works
  • How we can use it
  • Collecting data
  • Sharing data

5 of 30

How Co-ML Works

We have talked about how machine learning works. We said it’s like teaching a robot by giving it lots of examples, helping it spot patterns, and correcting its mistakes.

We used an example of animals and made cards to make it simpler and easier to understand.

6 of 30

How Co-ML Works

Now, we’re going to use an app called Co-ML, made by Apple, to bring this to life!

This app will help us see how computers can learn to recognise shapes based on what we teach them.

7 of 30

How Co-ML Works

We will teach Co-ML to recognise 4 shapes…

Circle

Triangle

Square

Rectangle

8 of 30

How Co-ML Works

We know from our experience that the more Data we have the better the computer will be at identifying the shape. So we need lots of images of shapes.

9 of 30

Collecting Data

Our Data Set is going to be images of our 4 shapes. We need as many images of circles, squares, triangles and rectangles as possible!

Where can we get the images from?

What do we need to consider when choosing our images? (See next slide)

10 of 30

Collecting Data

Good Images should…

…have the shape in the center of the image.

…be in focus.

…not have other shapes in the image.

…have varied types of shapes.

11 of 30

Collecting Data

In pairs, collect 10 images of each shape.

You can use the internet or take pictures using your iPad camera.

Remember to capture ‘Good Images’ like we discussed earlier.

What might happen if our images aren’t ‘Good’?

12 of 30

Sharing Data

We should now have lots of images of shapes.

If we have all collected the right amount of images we should have over 500 images.

We need to get them all into one place.

To do this we will use AirDrop.

AirDrop

13 of 30

Sharing Data

We will use AirDrop to transfer your Images on to 1 iPad.

We will need to do this carefully, and 1 device/pair at a time.

I will let you know when to do it and help you.

Once we have all of the images on a single device we can start to train our Co-ML model.

AirDrop

14 of 30

Sharing Data

Open your Photos app and select all of your shape images.

AirDrop

15 of 30

Sharing Data

Tap on the ‘Share’ icon in the lower left corner, then AirDrop.

AirDrop

16 of 30

Sharing Data

Now wait for me to come around and transfer your images…

AirDrop

17 of 30

Sharing Data

Now we have all of our images on a single devices we can build and test our model in our next lesson.

Before we finish let’s test ourselves again on our Machine Learning Vocabulary.

18 of 30

Machine Learning Vocabulary - Can you explain them?

Machine Learning

Label

Features

Prediction

Training

Algorithm

Patterns

Data

19 of 30

Machine Learning Vocabulary

Definition: When computers learn from examples instead of being told exactly what to do.

Example: Like teaching a robot to recognize animals by showing it lots of pictures.

Machine Learning

20 of 30

Machine Learning Vocabulary

Definition: Information that computers use to learn.

Example: Pictures, numbers, words, or sounds.

Data

21 of 30

Machine Learning Vocabulary

Definition: Things that happen in the same way or look similar.

Example: Stripes on a zebra or the way birds always have wings.

Patterns

22 of 30

Machine Learning Vocabulary

Definition: A set of instructions that tells a computer how to solve a problem.

Example: Like a recipe that a computer follows to bake a cake.

Algorithm

23 of 30

Machine Learning Vocabulary

Definition: Teaching a computer by giving it lots of examples.

Example: Showing pictures of cats and dogs so it can learn to tell the difference.

Training

24 of 30

Machine Learning Vocabulary

Definition: A guess that the computer makes based on what it has learned.

Example: The computer guesses that a picture is of a cat because it has fur and whiskers.

Prediction

25 of 30

Machine Learning Vocabulary

Definition: The important parts of something that help the computer recognize it.

Example: A bird has feathers, wings, and a beak.

Features

26 of 30

Machine Learning Vocabulary

Definition: A name or category we give to data to help the computer learn.

Example: Labeling a picture of a dog with the word "dog."

Label

27 of 30

Machine Learning Vocabulary - Can you explain them?

Machine Learning

Label

Features

Prediction

Training

Algorithm

Patterns

Data

28 of 30

What have we learnt this session?

Understanding Co-ML

Machine Learning Vocabulary

Sharing our Data

Creating a Data Set

29 of 30

What will we be learning next session?

Next lesson we will be importing our Data Set in to the CO-ML and testing our model out! This will include…

  • Setting up a new Project in Co-ML
  • Creating Labels
  • Importing Data
  • Testing our model
  • Reflecting on our journey

30 of 30

Project Website

This presentation, and all project resources, are available on our project website…

www.TheBatterseaProject.net