Smart Tracking & Real-time Evaluation for Therapeutic Corrective Health

Two main parts of the project are:

• Classifier

  This part of the project uses a KNN Classifier to classify the yoga poses.
  Here we have trained the classified on a limited number of yoga poses. The poses that we have focued on are

  • Mountain Pose
  • Goddess Pose
  • Garland Pose
  • Plank Pose

• Correction system

  The Correction system uses angles between specific joints or body points, which are calculated from the body points given by posenet model. The algithm checks if the angles lie in the permisible range. If the practitioner performes yoga incorrectly then this is found from angles which are out of defined range and user is prompted accordinglt to correct his/her pose.