DubBetter Ring

Inspiration

We wanted to build an IoT device for users who care home surveillance, from both unwanted disturbances and third-party companies who have 24 hour access to your front porch, while also providing a seamless and intelligent way to notify users of critical events that may occur. We were inspired by the desire to combine convenience, safety, and smart home integration into a single device.

What it does

DubBetter Ring detects and recognizes visitors at your door using a camera connected to a Raspberry Pi 5. Once connected to your Wi-Fi, the user is able to load up our client and view all events and even the live camera feed.

How we built it

Our IoT device utilizes a Raspberry Pi 5, USB camera, and AI models to detect activities from the camera feed which then sends any unusual or important activity to our website with the use of web sockets to provide real-time alerts for our users.

Challenges we ran into

Training our AI was one of the more bigger challenges, along with display real-time data for our frontend client. The data sets we needed did not exist, so we had to record a few videos of our own to replicate doorbell situations, which took up time and still was not sufficient enough for a optimal AI model. Another issues we came into was with hardware limitations. The Raspberry Pi was used as our backend server connected to clients via web sockets, and data processing wasn't the fastest.

Accomplishments that we're proud of

We are proud to announce that although the project wasn't perfect, it was well designed. The AI model was able to detect situations even if it is only a few kilobytes. We were also able to upload real-time camera feedback/stream and managed to get our web sockets communication and working.

What we learned

We learned that when creating software, something is always bound to go wrong, but it's not about the issues you encounter, it's about how you recover and deal with the situation at hand.

What's next for DubBetter Ring

Our next goal is to implement a functional logging feature for events. The client so far is able to view events through test in the camera feedback, but it would be much nicer and cleaner if the user was able to view a list of events for what happened. We were also able to get SMS notifications for lour events, but did not manage to get far enough to display them correctly in text, so this functionality has been vaulted. Potential future development plans could include more hardware resources, such as a button to act as a doorbell, familiar facial recognition to show events for familiar guests, visitors, and overall just more memory to play around with.

Built With

• javascript
• python
• react
• tensorflow