Inspiration

This project was inspired in a way by both of our grandmothers. Specifically, Owen had a recent scare that led to a multi-week hospital visit. Eileen, his grandmother, has lots of health issues (diabetes, oxygen issues, etc). One day, a family friend stopped at Eileen's house and found her in a very confused state. She has her bathtub filling with water, gas burners on, sink water running, candles lit, and all just generally stuff that you don't do, then leave attended. The family friend quickly called 9-11. This is when she was taken to a hospital to find her pulse ox was in the 70s and her blood sugar was in the low 50s. Without the family friend stopping, it is very much in the air what would have happened. It begs the question: can we do something to help and monitor the elderly asynchronously to prevent them from harming themselves?

What it does

  • Movement tracking around the house (done with a combination of OpenCV, Yolo v8, and calibration squares to calculate real-time distance/movement)
  • Key Item detection (via a Roboflow model) and subsequent tracking of said items around the house
  • React / Vite front end UI
  • Backend database to store relevant information about locations and the time an item was last seen

How we built it

  • Frontend: React, Vite, Shadcn UI, Tailwind
  • Backend: FastAPI
  • Person Tracking: Yolo v8
  • Item Tracking: Roboflow
  • Camera: OpenCV

Challenges we ran into

  • MAP (Getting the icons to overlay correctly, narrowing down the tracking to be more accurate)
  • UI/UX. Limited experience with designing interfaces.
  • Crawling around the floor to tape squares to the ground is painful
  • Greyson got sick during hacking hours

Accomplishments that we're proud of

  • Getting it to work
  • Tracking system +- 2inches
  • First ever hackathon submission for Greyson

What we learned

  • Anything and everything about Cameras. Similar to LLMs but very different. Great learning experience.
  • More experience with React. Very limited experience beforehand
  • How to create a MVP within a short timespan

What's next for Granny Guard

  • Detection of stove being turned on (via Apriltags)
  • Detection of doors being unlocked (via Apriltags)
  • Detection of running water (via a greyscale mask & motion detector)
  • Smoothing feature for the map that would prevent sudden jumps in movement
  • Face detection (with features from OpenCV)
  • Fall detection (via OpenPose or MediaPipe)
  • More personalization (having cameras be in custom designs, more personalization on the elderly person's dashboard, etc.

Built With

Share this project:

Updates