Text recognition with LearningML and Scratch

Welcome everybody to the tutorial Text Recognition with LearningML and Scratch. 00:00:00

We start by navigating to the LearningML platform. So last time, I saved the project to my account. 00:00:05

So I can now easily retrieve it by going to My Things and loading the projects. 00:00:11

So this way, I don't need to create the classes again and add all the training examples. They're 00:00:18

already there. You can also upload projects that are saved on your computer. So let's train the 00:00:23

model and once it has trained let's go to the programming platform by clicking on the scratch 00:00:28

icon okay so we're now in the scratch programming platform to which we have added our own blocks so 00:00:35

today we will work with the learning ml text blocks so let's let's try it so what did we 00:00:44

want to do well we want to have a virtual assistant um let's add a costume for our 00:00:51

virtual assistant. So let's search for robot. Let's pick the retro one. Has it 00:00:55

been added? Yes. And let's remove the two others. Let's add some code for our robot. 00:01:07

So we want to have a nice robot, so it should talk to us. Hi there, I am your 00:01:13

assistant and then it should ask us what we want him to do and we can find this 00:01:24

in the light blue colored blocks so he asks us a question how can I help you 00:01:33

and let's also add an event to start this sequence of blocks okay so let's 00:01:43

click on the flag and our robot says how asks how can I help you and a prompt 00:01:51

appears so let me write there lights off and nothing happens that is because 00:01:59

we're not telling the robot to do something with that answer so let's 00:02:08

let's add that so we have here the block answer so that is what we just that is 00:02:16

our input. So, for now, let's start with something easy. Let him just repeat the answer. 00:02:23

This. And let's try again. So, the prompt appears. Lights off. And he just repeats our answer. 00:02:30

Okay. But we don't want him just to repeat the answer. We want him to understand what we want. 00:02:44

So let's use our AI model. 00:02:50

And we have a block that says classify text. 00:02:54

So it takes a parameter text. 00:02:57

So that is our input, our answer. 00:02:59

So let's move answer there. 00:03:01

And let him repeat this. 00:03:06

Lights off. 00:03:13

And now he repeats switch off. 00:03:15

So the label, the predicted label for our input is switched off. 00:03:19

and our robot has repeated that prediction that label okay so but that's not everything 00:03:24

we also want him to do something so let's add a backdrop so we can actually switch a light off 00:03:34

and on so let's find one um let's take a bedroom this one let's move our robot a bit here okay 00:03:40

let's go back to the backdrop let's give it a name so this is the one where we have the lights 00:03:51

on and then we have another one um let's rename this as lights off and let's hack it a bit 00:03:58

um so let's add a rectangle here a bit clumsy like this and let's make it dark 00:04:08

and let's call this lights off. So, let's add some code for our backdrops. 00:04:22

So, let's think of it first. So, what needs to happen is the following. So, when the robot 00:04:33

receives an order, it should send a message to the current backdrop to change. If the message 00:04:38

is switch on, the current backdrop should change to the lights on backdrop and otherwise, when it 00:04:45

receives switch off the current backdrop should change to the lights off backdrop so how are we 00:04:51

going to do that well we have other blocks for that so we have event blocks when i receive 00:04:57

a message so we're now adding code to the backdrop 00:05:05

so let's add a new type of message the first one switch on and we have a second message 00:05:10

We have two types of message. 00:05:19

One is switch on, the other one is switch off. 00:05:22

So we have two messages. 00:05:27

What do we need to do? 00:05:28

We need to change the backdrop. 00:05:31

So we need to change the looks. 00:05:33

So that's in the purple blocks. 00:05:35

So what needs to happen here? 00:05:37

Well, when the message is switch on, the backdrop should change to the lights on backdrop. 00:05:38

When the message is switched off, the backdrop should change to the lights off backdrop. 00:05:46

Okay, so that's the code for our backdrop. 00:05:52

Let's now go back to our robot. 00:05:56

So let's run the program just once more, see what happens. 00:06:01

So off, let's be really short now, switch off. 00:06:07

So nothing has changed because the backdrops, they haven't received any message. 00:06:11

So the robot needs to broadcast the message and for this we have other event blocks. 00:06:16

So the robot needs to broadcast the message it has just received. 00:06:26

So let's add another classified text together with the answer. 00:06:33

So now it will just repeat, it will say what it will do and then it will broadcast that 00:06:44

message, such that the backdrops, the current backdrop will change. So, hi there, I'm your 00:06:49

assistant. So, lights off, switch off, and the backdrop has changed, and the lights are off. 00:06:56

So, let's switch them back on. So, switch on, and the light is back. So, this is just a very simple 00:07:06

illustration of how you can use the AI models within Scratch. Other examples can be found 00:07:20

on the LearningML webpage and our FIAS project page. See you in a next tutorial! 00:07:26

Text recognition with LearningML and Scratch - Contenido educativo

Del mismo autor…

1TRIMESTRE INFANTILD

Clase11_Aceleración

¿Dónde vas con mantón de Manila?

Una noche en el museo de Ciencias Naturales

Pasos de chotis

Cambia los fondos con Scratch para hacer una portada cuando programes un juego

El baile de Santa