REAssist

Your third eye to keep track of your surroundings.

Comment

Ever felt bored in class and wondered why? The answer may be simple. You're simply too cool for everyone.

Instead of constantly being with a group of bums, REAssist acts as your third eye, monitoring the room environment. Maybe the reason nobody wants to approach you wasn't actually you, it was those around you. With REAssist, you can now track the rating of the room you're in, and when the score drops too low, it might just be a good reason to leave. Don't associate yourself with people that bring you down. You deserve better.

On the other hand, REAssist also notifies you when someone with a high score enters the room so you won't ever be missing out. With REAssist, you'll always get the first mover's advantage on that one perfect guy or girl you'll never speak to anyway.

Building

REAssist is build entirely on Python, utilizing tools such as DeepFace and OpenCV for facial recognition. The individual and room scores are computed based certain algorithms, primarily the Golden Ratio Face. The program then uses the a Discord bot to send direct messages as a notification whenever a certain threshold is reached.

Challenges

One of the major limitations of REAssist is the conditions it can work in. The bot uses the Apple Camera and must be connected to a Mac (can be wirelessly). Additionally, there were many compatibility issues with computers and libraries. Much of the time was spent attempting the install the libraries correctly, rather than actually using them. The bot also kept detecting water bottles as faces, and for a long time we thought this was a problem with the algorithm. This ended up being a cropping issue, where the bot wasn't taking proper screenshots of the inputted video footage from the phone.

All jokes aside, building this project was an extremely fun and rewarding experience. We learned a lot about implementing libraries, image recognition technology, linking multiple platforms and tools together, and of course, how to prompt GPT accurately.

Built With

Share this project:

Updates