Workout helper app

The application is based on ML Kit Pose Detection API - a model for the real-time pose estimation in video streams and static images. The algorithm defines a pose as a set of points, located at specific parts of a human body such as shoulders, hips, etc. The varieties in the locations of these points allow for the differentiation between the poses.

Our developers have analyzed numerous images and videos, demonstrating the proper performance of exercies, in order to define how the coordinates of the points correlate with each other and the surrounding surfaces.

As a result, the app can detect improper positions of the human body parts during the performance of the 5 exercise types:

- Plank;

- Push-ups;

- Squats;

- Pull-ups;

- Lunges.

To estimate the body positions while working out, the user only needs to launch the camera within the app and set the phone so that the whole body is visible. The next step will be to select an exercise from the list, and you can start working out!

In case the exercise is being performed in an improper way, the mislocated body part will be highlighted, and a recommendation message will describe how to achieve the proper pose.

The initial version of the algorithm was developed in Python, and then Java was used to build an Android app.