Weight lifting is(, in the opinion of Daniel,) one of the most fulfilling sports, both psychologically and physically.
Many are scared of this sport due to horror stories of slip disk, hernias, broken spines or other bodily horrors. Despite the fact that weight lifting is relatively safe when compared to other sports like (American) football, basketball, or running.
One of the largest barriers to entry for those seeking to enter is the difficulty in developing proper and safe technique. Minor visible and invisible influence our biomechanics and physiology means there is never a one-size-fits-all way of performing any of the compound movements. In order to develop a sensor for proper form, one must either must be taught, or through possibly dangerous trials and tribulation, develop it from scratch.
PostureCheck seeks to ease the learning curve and promote safe and healthy wight lifting. We ahieve by providing immediate feedback of your exercises and diagnosis root causes of bad form through Pose detection model and internet-scale pre-trained LLMs. The pose detection monitors your limbs and complies notes that the LLM uses to diagnose and prescribe fixes to the root causes of improper form.
To achieve this, we used Google's Mediapipe and OpenCV to process both live and per-recorded videos to track landmarks of the the users limbs. This data is used to calculate various and identify various traits such as limb length, checking if knees went over toes, caving knees, or other important signal good or bad movements. These are traits are then feed into a LLM to both evaluate the overall movement, highlight areas of growth, and recommend actions to address any issues found.
For our front end, a simple HTML, CSS and Node JS provided a simple front end for users to watch their movement with greater insight. This also serves as the interface for user to ask follow up questions to the model.
Log in or sign up for Devpost to join the conversation.