FitScroller

• 

  In Progress

• 

  Before Connection

• 

  Tracking

Inspiration

To combat social media addiction (brainrot) by gamifying fitness. We wanted to make sports more engaging by turning Instagram Reels into a workout experience.

What it does

Complete a push-up to scroll to the next Reel. Blink to like a reel.

How we built it

Backend: Python + OpenCV (Camera processing) + MediaPipe (Pose/Face tracking) + PyAutoGUI (Input simulation) Communication: Flask server acting as a local API bridge Frontend: Chrome Extension (JS/CSS) for HUD injection and real-time data fetching

Challenges we ran into

Security issues prevented access to camera data in Chrome. Hence we had to use a separate python server to obtain those input and process them.

Noisy data for blinking detection. To improve it we kept a buffer of readings and modelled it as a Normal Distribution to test if the newest reading was "significant" (counted as blinking).

Accomplishments that we're proud of

Cool UI. Detecting location of like icon instead of hard coding it to allow for more portability.

What we learned

Push-Ups can be very tiring, especially when you have to watch a whole reel before doing the next one.

What's next for FitScroller

• Including other exercise moves as well eg: squats.
• Fine-tuning the detection model for blinking.

Built With

• chrome
• css
• flask
• javascript
• python