Foundation allows users to receive feedback on their lifts without the need for a personal trainer
Many gym-goers want to improve their lifting technique, but many either don’t want to deal with the risk of self-teaching, the cost and scheduling hassle of hiring a personal trainer, or the clunkiness and delays associated with consulting an online trainer. Unfortunately, in order to get the results they seek, they must succumb to one of these options. In order to remedy this problem, our team has been developing Foundation, an app that records the user performing their lift, detects any flaws within the lift, then critiques the user so that they may improve their form on the next attempt.
The way that Foundation works is it takes in a video that the user inputs, detects the type of lift that user is trying to perform, and then maps the movement of their joints into sets of 2D keypoints, separated frame-by-frame. From this data, the app utilizes algorithms trained on over 12,000 labeled videos sourced from Instagram to detect the user’s flaws. After a flaw has been detected, Foundation gives feedback to the user in real-time so they may correct their form.
https://github.com/d3nnyhl/foundation