This website analyzes movement and voice in real time to create artificial intelligence

teachable machine is a tool based on the Web this makes it possible to quickly create machine learning models, which teachers and students have used to explain different topics in a more interactive alternative. Moreover, it is compatible with Google Drive.

This website is for anyone with an idea to explore, such as artists, students, innovators or creators of all kinds. Also, no prior knowledge of machine learning is required. Here you can create unique commands that use artificial intelligence to identify any sound, movement or person.

On this website, you collect and group the examples you want the computer to learn by classes or categories, where you can prepare the model and test it on the fly to see if it can correctly classify the new examples. Likewise, these projects can be exported for sites, apps and more. The template can be downloaded or hosted online.

You may be interested: Truth or lie: putting magnets in the fridge uses more energy

The Teachable Machine uses files or captures live examples. In addition, according to the entity, “it is a tool that respects the way of working, and can be used on the device, without any data from the webcam or the microphone leaving the site”

The models created with the platform are TensorFlow.js that work anywhere with javascript. Therefore, they are compatible with tools such as Issue, P5.js, Node.js and many more.

Also, you can export the templates in different formats to use them on other platforms.

You may be interested: Virtual reality would affect vision for this reason

Pose detection means that from an image of a person you can estimate the pose, for example the location of the arms, legs, head and joints. To start training the machine, you must first create different categories, or classes, to teach it.

There you can do different classes like “No Tilt”, “Left Tilt” and another one called “Right Tilt”. Later, example poses can be added for each of these classes.

Also, if you need to record samples without holding down the button, you can go to the settings panel and disable Hold to Record, and it will count down and then automatically record a set of samples.

You may be interested: What is spyware and how to know if there is one on the cell phone

And then you can click on train and once the model is finished you can see the result of the poses that have been made.

There is also a machine learning model to detect clicks, claps and hisses using audio clips. To use it, you must first create different categories, or classes.

The platform offers two checkboxes: “Background noise” e “class 2″. These are the initial classes that Teachable Machine handles when running any project. This will allow you to create sound commands with people who cannot speak, to identify what they are asking at the time.

You may be interested: How to take advantage of Meta’s new artificial intelligence to do research

To use this you will need a background noise class to detect when no noise is being produced. And because background noise in a forest is different than in an office (or anywhere else), you should give this class audio samples for wherever you plan to use your model.

This particular machine learning technology learns from one-second samples. Classes need at least 8 one-second sound samples to practice properly. The more classes of data to learn, the better they will be ranked.