Inspiration
We noticed that many people, especially students and office workers, spend long hours sitting in front of computers. This often leads to bad posture, back pain, and long-term health issues. Our inspiration was to create a simple, AI-based solution that helps people maintain correct posture and avoid health problems.
What it does
1.Detects human body posture in real-time using a webcam.
2.Identifies whether posture is correct or incorrect.
3.Gives instant feedback (visual or audio alert) to help the user correct themselves.
4.Can be used during work, study, or exercise sessions.
How we built it
1.Used Python, OpenCV, and MediaPipe Pose for pose detection.
2.Extracted body landmarks (shoulders, hips, spine, neck) to calculate angles.
3.Applied simple rule-based conditions (e.g., if the back angle deviates more than 15° = slouching).
4.Built a real-time feedback system to alert the user when posture is wrong.
Challenges we ran into
->Difficulty in tuning the angle thresholds for different body types.
->Making the model work in real time without lag.
->Handling cases where the user moves too much or the camera angle is not perfect.
Accomplishments that we're proud of
->Successfully implemented real-time posture detection.
->Built a working prototype that gives live feedback using a webcam.
->Made the project simple and lightweight so it runs on normal laptops.
What we learned
->How to use pose estimation libraries like MediaPipe and OpenCV.
->Importance of data preprocessing and setting correct rules/thresholds.
->How AI can be applied in health and fitness domains to solve real-world problems.
What's next for Body Posture correction model
->Develop a mobile app version using TensorFlow Lite.
->Add daily posture tracking reports and analytics for users.
->Use deep learning classification (CNN/ML models) instead of only rule-based logic.
->Integrate with wearable sensors/IoT devices for more accurate correction.
