DoorAuth

As you can see if the current image matches when we run test client it will compare with authorize.jpg and say if the auth is successfull.

Inspiration

What it does

DoorAuth is a real-time access guard:

A sensor detects a possible “open” event (we used a light-based trigger for the demo).
A laptop webcam captures a live image and compares it to an authorized reference photo using Gemini vision.
The system returns a strict true/false decision:
- AUTH_SUCCESS → no alarm
- AUTH_FAIL (unknown / not logged in / unsure) → buzzer alarm
It’s designed to fail safe: when in doubt, it alerts.

How we built it## Inspiration

We wanted a simple, fail-safe way to deter unauthorized access to a box/door (ex: dorm room storage, lab cabinet, shared workspace). If the system can’t confidently verify the authorized person, it should default to “deny” and alert.

What it does

DoorAuth is a real-time access guard:

A sensor detects a possible “open” event (we used a light-based trigger for the demo).
A laptop webcam captures a live image and compares it to an authorized reference photo using Gemini vision.
The system returns a strict true/false decision:
- AUTH_SUCCESS → no alarm
- AUTH_FAIL (unknown / not logged in / unsure) → buzzer alarm
It’s designed to fail safe: when in doubt, it alerts.

How we built it

Python + OpenCV to capture images from the webcam and encode frames.
Gemini (google-genai) to compare the authorized reference image vs the live login image and return JSON.
Socket networking (TCP) so a device (e.g., Raspberry Pi) can request authentication from the laptop server.
A real-time loop on the device triggers the buzzer based on the auth response.

Challenges

Rate limits and reliability: we added retry/backoff for API limits and made the decision stricter (only accept a match when confidence is high).
Networking: making sure the device and server can communicate reliably on the same network.
Fail-safe behavior: ensuring the alarm triggers by default if auth can’t be confirmed.

What we learned

How to wire a real-time sensor trigger into a secure “verify then act” pipeline.
How to structure a system with clear separation: camera/auth server vs device alarm loop.
Practical debugging of networking + real-time loops under hackathon time pressure.