This project deals with 2 issues:
-
Use of openCV and Mediapipe to first gain body landmark coordinates on a projected on a 2-d surface.
-
Exploration of 2 possible applications of having this landmark data:
- Imgage classification(pose estimation folder): For Image classification, I have used a supervised data of classified images and trained a neural network classification model for possible identification of various yoga poses using relative positions of body landmarks from the available images.
- Reps counting for a sample excercise(Workout helper folder): Since the landmark data can be captured for real time videos as well, I have attempted to count the reps for a sample excercise based on the changes in landmarks relative positions.