Detection of Yogasana Postures

Team: United Tots

This project is the winner of Second Runner Up of Codeavour 2020 AI

Project Details

Team Details

  • Member 1: Yashwitha Balasubramaniam
  • Member 2: Michelle Dominic
  • Age Group: group_2_(10_years_-_13_years):_participants_falling_in_this_group_must_have_been_born_between_2_october_2007_and_1_october_2010.
  • Country: India
  • School Name: san_academy_group_of_schools
  • Partner: stempedia

Project Explaination

Our World in now facing a severe crisis of the Pandemic Covid-19. This has affected the lives of many beings and has created a larger impact all over the countries both financially and economically. The stress level of the humans has increased to the core and there is big drop in their immune level as well.

To handle this tough phase ,practicing of Yogasanas regularly will help us to stay fit and healthy.

We have chosen the project under the Theme- Beat the Pandemic with AI (Artificial Intelligence) using the extension of  MI (Machine Learning) Component in Pictoblox.



  • We have created around 250 Image Samples of each posture to classify an asana.
  • We use these image samples to train our model.
  • Export the trained model to get the hyperlink which is then copied and loaded in to the Pictoblox.
  • Then comes the Coding part.


  • Fetch live feed from Web camera
  • Process the feed with trained classes
  • We look for 95% Confidence level (accuracy) of postures
  • System identifies the postures based on the accuracy level


We have achieved the expected accuracy level in the detection of different postures of asanas using the Machine Learning Extension .

We have trained four different postures of asanas which are closely related to strengthen the lungs and boost our immune system. This project can be extended to include more asanas . We can have both fun in learning programming and learning different asanas as well.