PowerRecon

Dashboard
GitHub
PowerRecon

Inspiration

The solution to this Hackathon, PowerRecon, was pitched by one of our teammates, Ting Xiao, he realises that the library in the school is unnecessarily cold and people usually don’t turn off the appliances when they leave the classroom. He recalled the lecturer mentioning to him that the power market in the industry is a direction worth exploring as there are currently not many advanced technologies being implemented and companies are losing profits solely from the added cost of unused utility consumption.

What it does

PowerRecon is an object presence detection camera that uses computer vision models to track particular subjects in a room and intelligently adjust the appliances inside the room. PowerRecon is able to distribute power in the room intelligently.

How we built it

The current prototype uses the laptop webcam to act as the IoT camera unit that will feed the live video stream into the object detection model using python OpenCV. The model that we used for doing the object detection was Google’s Mediapipe and the output of the model is sent directly to the node server via Socket.IO connections. The Front-End dashboard uses the same Socket.IO protocol to receive the information sent from the camera unit through the server. The end product was then fully hosted on Heroku.

Challenges we ran into

During the process of building the prototype, we faced many challenges ranging from selecting the solution that we want to implement to choosing the right model for our use case. These challenges took up a majority of our time as we were unsure of how we should start building our solution prototype.

Accomplishments that we're proud of

We are very proud of being able to set up a deployable prototype that allows users to test and play around with the object detection feature as well as the dashboard of the solution.

What we learned

Through this hackathon, we can safely say that we were able to learn much new stuff that we have never used before such as using Socket.IO for communicating between the python model and the node server and utilizing Google MediaPipe API for object detection.