A simple browser-based posture detection app that uses TensorFlow.js and PoseNet to give real-time feedback on your sitting or standing posture via webcam. Built during my first year of BTech CSE to explore machine learning in web development. Posture Alert* is a browser-based application that leverages machine learning to monitor and assess user posture in real time using a webcam. Built using TensorFlow.js and the PoseNet model, it provides instant visual feedback to help users maintain good sitting or standing posture.
- Real-time posture detection via webcam
- Feedback messages such as:
- "Good posture"
- "Sit straight!"
- "You're slouching"
- Lightweight and installation-free — runs directly in the browser
- Designed with simplicity and accessibility in mind
- The application accesses the user's webcam and processes video frames in real time.
- PoseNet identifies key body points such as shoulders and hips.
- Based on the relative vertical positions of these points, the app determines posture quality.
- Feedback updates continuously to encourage proper alignment.
- TensorFlow.js – Browser-based machine learning
- PoseNet – Real-time human pose estimation
- HTML5 and JavaScript – Frontend logic and rendering
- Canvas API – Drawing keypoints and live video frames