Inspiration
Many people are sensitive to noise and crowding. Others prefer lively, busy spaces to study with peers. While we all have our niche study spots, sometimes even those rooms are too full, or too quiet, for our needs. We decided to build a solution using sensors and crowdsourced data, while keeping privacy at the forefront and retaining no recordings of students on external servers.
What it does
Our project combines sensor recording & processing via computer vision with a web frontend in order for students to quickly be able to search for rooms matching their desired population and noise level. A webcam linked to a controller (in our case a laptop) takes photos and an audio recording every minute, updating the live website.
How we built it
Every minute, photos from the webcam are scanned by a small computer vision model to count the number of students, and the audio recording is processed to produce an average noise level. The recordings are deleted and no data other than the person count and average noise level are retained. These are then written to the Tiger Data database backend, storing live room readings. The Vercel frontend reads these dynamically, providing the user with instant information.
Challenges we ran into
Some of the main issues we ran into were linked to the hardware we were trying to use, in our case, the Arduinos and Raspberry Pi. After trying MLH's entire stock of microcontrollers, we still couldn't get any of the RPI4s to boot. As we initially planned for the project use Raspberry Pis signal input, we had to think on our feet. We ultimately decided to prototype our project with laptops and to add an option to crowdsource user data.
Built With
- arduino
- claude
- codex
- javascript
- python
- raspberry-pi
- tiger-data
