00:01Here is the easiest way to add AI capabilities to any robot built with the BBC micro:bit

00:06If you want to learn how to create a basic AI application with image or sound recognition, there is no better way to start than playing with Teachable Machine app from Google

00:16You open the online app, grant access to your webcam, you grab and upload a few images of your hands or something else, click train model, and voila, you have an AI trained to recognize your hand gestures running on your computer

00:33The same goes with any object, sounds, body postures, or whatever

00:35And what could be even cooler if you could take the model you just Trained and connect it with your DIY robot built with BBC micro:bit, and now you can with my mini app, Teachable Micro:bit, available online on Cardboard Robots website

00:52When you export your Teachable Machine model to Google Cloud and open a sharable link in the Teachable Micro:bit app on your computer or smartphone, you can then connect it with the BBC micro:bit and send data from the AI model to your robot via Bluetooth

01:11Of course, you can program your robot to do whatever you like, so what can you build with it? Maybe a color sorting robot that can sort ping pong balls, or an intelligent trash can that can sort paper and Plastic, or a line follower robot, or automatic gate for Hot Wheels cars, you could also control your robot with hand gestures

01:35What is the best in this app is that you can open it on your smartphone and then mount the smartphone on the robot, so it is very easy to build something like an robotic car with AI camera without any specialized sensors or electronics

01:48You only need a Micro:Bit and a smartphone

01:50If you are a teacher and you want to introduce your students to AI and machine learning concepts, I cannot think of any more engaging tool than doing this by building intelligent robots

02:03You can use examples shown

02:05On cardboard robot's website or invent your own

02:07So how does teachable microbit app work? This app is a simple bridge that allows you to connect any model trained in teachable machine with any robot built based on BBC microbit

02:21It works as a progressive web app in your browser on desktop or mobile devices

02:27Training the model, you do all of the AI training in the teachable machine app

02:33When your model is ready and works good in the preview window, you can export it to Google cloud, and you will receive a sharable link to use this model in other apps

02:45Coding the robot, on the app website, you can find template code for Make code, you can modify it to add any logic you need to your robot

02:54Micro:bit will receive simple text commands with detected class names, so we can build the logic of your robot with simple if conditions

03:01For example, if received string is "okay," then show happy face on your Micro:bit display

03:10If received string is "not okay," then show sad face on your Micro:bit display

03:14Also, make sure that your project settings are set to "just works," otherwise the app will not detect the Micro:bit for Bluetooth connection

03:25Connecting AI model and Micro:bit, once your robot is ready, open Teachable Micro:bit app on your computer Chrome web

03:29browser, or your smartphone

03:33On iOS, you have to use Bluefy web browser

03:37You can find it in the App Store

03:40When you click start, the app will ask you for a link to your model

03:44Grant permission to use the camera, or your microphone, and when it opens, you will see labels of detected classes below the camera image

03:51Click the robot icon in the upper right corner and choose your Micro:bit board from the pop-up window to connect via Bluetooth

03:59If everything is okay, you will see a round outline on the robot icon, and detected class names will be sent to the Micro:bit via Bluetooth

04:09If you coded the Micro:bit properly, your AI robot should

04:12Be finished

04:14There is one caveat you have to be aware of: when using Teachable Machine and the Teachable Micro:Bit app, when you build a robot with machine learning capabilities, you have to put a lot of attention into the model training phase

04:28The quality of your model depends on the quality of the data you will give it

04:31As always, it will probably require a lot of trial and error, and you will have to update the model a few times until it works as intended

04:40And now pay attention: when you export the model to the cloud and open it in the Teachable Micro:Bit app, it will be cached in your browser's memory, and even when you update the Model in Teachable Machine, the app will still be using the old cached version

04:57To be sure that the Teachable Micro:bit app uses the updated version of your model, you can open the app in a new incognito tab

05:09Reusing the models, the app does not have a regular save function to store and access links to your models

05:16If you want to reuse your model and do not want to paste the link manually every time, you can pass it as a parameter in the URL address of the app

05:27This way, you can simply store and access multiple models as regular website links

05:32Remote camera, as I said before, one of the biggest features of this app is that you

05:38You can launch the train model on a smartphone and mount the smartphone onto your robot, but there is one problem with model training, because Teachable Machine has no mobile version and works only on desktop computers with built-in webcam, and in specific cases, it is important to train the model with particular view from the camera it will be used on, so it would be best to train the model with the view from your smartphone camera

06:05The workaround for this problem I can suggest is to use a camera sharing app, such as KMO, to share your smartphone camera to your computer for the training process, and that is all

06:19Always, I hope that you will find this app useful and get some valuable learning experience from it, and if you build something interesting, please let me know

06:30Cheers!