Inspiration
We recently set up a security camera in our college apartment after getting a server to host our website. It worked fine, but I quickly realized its biggest limitation — it could only monitor one room. I didn’t want to spend more money on multiple cameras, so I started thinking: what if we used a drone instead? That’s how the idea for a drone-based security system was born — one that can track intruders, follow them around the apartment, and even snap pictures in real time.
What it does
Our system uses facial recognition to detect and track a person’s face. Once an intruder is spotted, the drone automatically follows them and records their movement. It can also take snapshots of the person and save them for review later — all while navigating around the space autonomously.
How we built it
We combined off-the-shelf facial recognition libraries with a basic drone SDK. The drone takes camera input, processes it on-device (or through a lightweight server), and uses that to make real-time movement decisions. A lot of testing went into making sure the drone could actually keep up with someone moving and respond quickly enough.
Challenges we ran into
Facial tracking was no joke — it took a lot of tweaking to get it to work consistently, especially when lighting or angles changed. Getting the drone to fly smoothly without crashing while also tracking a moving target was another huge pain point.
Accomplishments that we're proud of
Honestly, the fact that we got it working at all is something we’re super proud of. Watching the drone actually follow someone and take photos was one of the coolest moments — it felt like something out of a sci-fi movie.
What we learned
We learned a ton about real-time computer vision, how to control drones through code, and the limits of consumer tech when you try to push it just a little further. We also realized that hardware projects are a different kind of grind — more bugs, more crashes, but way more satisfying when things click.
What's next for MarlowOS
We want to make it smarter — like recognizing specific people, integrating voice commands, or syncing it with a full smart home setup. We’re also thinking about turning this into a more accessible plug-and-play security solution that anyone can set up at home without needing multiple expensive cameras.
Log in or sign up for Devpost to join the conversation.