PostureGuard AI

Inspiration

We spend hours a day hunched over laptops — studying, coding, gaming — and the damage happens so slowly we don't notice it until our neck aches or our eyes burn. The problem isn't that people don't care about posture and eye strain, it's that there's no feedback loop. You can't feel your spine curving forward in real time, and you definitely can't feel yourself blinking less. We wanted to build something that closes that loop using hardware everyone already owns: the webcam built into every laptop.

What it does

PostureGuard AI turns any laptop webcam into a real-time ergonomics coach. It uses computer vision to:

• Track posture — monitoring neck and shoulder angle to detect forward-head slouching as it happens
• Track blink rate — detecting when blinking slows down, a key sign of digital eye strain
• Alert instantly — nudging the user the moment either metric crosses a healthy threshold, before bad posture or dry eyes turn into real pain
• Show session stats — giving a simple summary of how much time was spent in good posture and average blink rate, so users can see progress over time

No wearables. No extra sensors. Just the camera already sitting on top of the screen.

How we built it

We built PostureGuard AI in Python using OpenCV for webcam capture and Google's MediaPipe library for pose estimation and face mesh tracking. Pose landmarks (ears, shoulders) are used to calculate a neck-forward angle in real time; if that angle passes a slouch threshold for more than a few seconds, an alert fires. Face mesh landmarks around the eyes are used to calculate eye-aspect-ratio, which lets us detect blinks and calculate blink rate per minute without any specialized eye-tracking hardware. [Add your dashboard/UI framework here, e.g. Streamlit/Tkinter/simple HTML overlay], and alerts are delivered as [describe: popup / sound / on-screen banner].

Challenges we ran into

[Fill in honestly once built — e.g. "Tuning the slouch-angle threshold was harder than expected: leaning in to type looked identical to bad posture at first, so we had to add a time-based tolerance instead of an instant trigger." Specific, real challenges score much better with judges than generic ones.]

Accomplishments that we're proud of

• Built a working real-time computer vision pipeline solo during the hackathon
• No hardware beyond a laptop webcam required — fully accessible, zero cost to the end user
• Turned two silent, hard-to-notice health problems (posture and eye strain) into something visible and actionable in real time

What we learned

[Fill in once built — e.g. what you learned about MediaPipe, angle math, computer vision thresholds, or scope management during a hackathon.]

What's next for PostureGuard AI

• Add posture/blink trend graphs over multiple sessions so users can track long-term improvement
• Personalized thresholds — calibrate to each user's natural sitting posture instead of one-size-fits-all angles
• Package as a lightweight background app (menu bar / system tray) instead of a foreground window
• Explore integration with break-reminder apps and calendar tools for smarter, context-aware nudges

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for PostureGuard AI

Built With

• ai
• computer-vision
• ergonomics
• eye-strain
• hackathon
• health-tech
• machine-learning
• mediapipe
• opencv
• pose-estimation
• posture-detection
• python
• real-time
• webcam