Final Prototype
MOTION DETECTION - 2D Camera Detection
Many different techniques for motion detection were considered when we are developing this part. 2D camera detection is our last choice, because of its excellent performance and high efficiency.
2D camera detection is based on a public machine learning CV model, MediaPipe, which is well-trained for tracking motion and construct motion maps in 2D coordinates. Though the 2D coordinate system may result in a slight error compared with 3D camera detection, it shares high working efficiency when analyzing massive data.
Thanks to MediaPipe (https://google.github.io/mediapipe/) for open source 2D motion detection library.
POSE CORRECTION
Limited by development time, we only implemented 3 types of exercises in this project - squat, push-up, hip-bridge. The algorithm can automatically count numbers and offer correction suggestions by measuring important joints and movement distance in pre-built tracking maps.
For the squat, we will measure the hip joint to check whether or not users did the exercise deep enough. And the distance from knee relative to toes are also considered to avoid any safety accidents.
For push-up, the arm joint is what we measured to make sure that this training is efficient.
For hip bridge, we are to measure the time users hold at top.
INTERACTIVE DESIGN
Our interaction design mainly turned to the UI design in the video on the right hand side, showing not only our whole game design but also how the reward system works and what can it provide to users.
​
​
​
​
​
Our game is to provide users a feeling of traveling on the world with the miles they get from everyday exercises. After visiting one place by using miles, the user may have a look of the city’s basic information and famous things, giving them the real sense of visiting and learning knowledge.
​
​
​
​
​
The distance in the game (in miles) is scaled down from the real distance between cities, which give users the real sense of traveling from far and close places and give them a basic idea about the distances between places in daily life.
APP INTEGRATION