Chitti app logo

Chitti is better in the app

Coins, streaks and builds — free on Android & iOS

Get the app
Chitti logoChitti

Sit and Stand

3 Minutes · Module 6 - Pose Processing

Learn pose machine learning for kids by training a Teachable Machine model to recognise sitting and standing poses. A Tamil technology lesson.

  • How games track body joints using stick-figure pose models
  • Uploading sit and stand sample images into Teachable Machine
  • Testing the trained model using a webcam and downloaded image
What this lesson teachesRead the lesson as text

A pose machine learning model is a computer program that recognizes human body postures by tracking the positions of joints. It allows software to detect whether a person is performing a specific physical action, such as sitting down or standing up.

How pose tracking works

Pose models operate by simplifying the human body into a stick-figure representation. Instead of analyzing full physical details, the model focuses on the location of body joints and how those joints are positioned. Gaming systems like the Xbox rely on this pose machine learning concept. When a player performs an action with their hands or body, the system tracks their joints so the character on the monitor moves in response.

Setting up classes and samples

  • A user opens Teachable Machine and creates a new pose project.
  • Two classes, which are categories used to sort data, are set up and named stand and sit.
  • Posture samples can be collected directly using a live webcam or uploaded from saved files.
  • Sixteen posture samples are uploaded for stand, and nineteen posture samples are uploaded for sit.
  • Clicking the train model button teaches the computer to recognize the differences between the two poses.

Testing the model

After the training process finishes, the user tests the model to confirm that it works properly. When sitting in front of the webcam, the model reads the user's body joints and displays an output of sit at one hundred percent accuracy. The input can also be changed from the webcam to file mode. Searching for a standing image in a web browser, downloading the picture, and uploading it into the model produces an output of stand at one hundred percent accuracy.

Using trained models

Training and testing a pose model is the first part of building machine learning systems. Once a model successfully learns how to distinguish between sitting and standing postures, it can be exported to create real computer applications that respond to human movements.

Part of the course

A.I. with Teachable Machine

View course →