Posture Detection System with Image Recognition
3rd Place Makethon Winner, University of California Riverside
With this project, we won the 3rd Place Winner at the Makethon BearHack, held at the University of California Riverside.
Problem & Solution
We tackled a problem we suffer as students: back pain after long study sessions due to incorrect posture.
We propose a real-time image detection system that alerts the user when their posture is incorrect.
Implementation
This project uses the MediaPipe library to create a real-time posture monitoring system using a webcam or video source.
Key nodes are placed on various parts of the neck and torso. Straight lines are then computed to connect these nodes. Posture is evaluated based on the angle formed by these lines and certain angles are classified as “good” or “bad.”
The script communicates with an Arduino. If the posture is detected as bad, a red LED lights up and a buzzer sounds to alert the user. When the posture is corrected, the LED turns green.
After each session, the user receives a report showing how much time was spent in good vs. bad posture, along with additional insights.