Our company applications are extinguished

Solve an everyday problem in six hours

Finding a way to help all colleagues in the company within six hours was not easy at all, because I knew that I wanted to try some new technology. And then it occurred to me that one thing we tackle in new offices every day: Reminding the last one to leave not to waste energy and turn off the heating, code, lock and, most importantly, turn off the lights.

Machine learning

I knew I would have to use machine learning, which had been enticing me for some time. The hardest part of it is getting quality data. However, a few photos of each light on and a few photos of the interior with the lights off were plenty enough to create the MVP app. My colleague Šárka Vrbovská was happy to collect this data. In the meantime, I started working on a simple application that shows the correct switch for the detected light in the plan of the offices.

Dataset 10 000 bilder

The created dataset for six lights and backgrounds, about 10,000 images in all, must have been sufficient. For training we used the Apple tool Creare MLwhich is very user-friendly and can generate a Machine Learning model that my application managed to use thanks to the Vision framework. (Create ML is only applicable to smaller datasets, which our MVP application met.) With the dataset set, the training took about two hours and we achieved 85% accuracy.

Turn off and run home

Using the app itself is simple: After taking a picture of the light you want to turn off, the app evaluates the photo and shows you where to look for the switch on the map. Once the lights are off, all you have to do is encode, lock and go home.

The expansion possibilities are endless

For example, you can improve the accuracy of the trained model or add an AR model of the office, which will easily guide the user/colleague to the right switch. Another extension could be the implementation of the user's location and office exit tracking, which would send a reminder to the latter whether he really thought about everything when leaving.

What other teams did at the Futured Hackathon:

⛺️ App to support sharing of business equipment, books, tents, tools or anything

📚 Expansion of the mobile toolkit with new functionality and the use of a completely new library for communication with the device

🛺 Our eco-wing worked on a Slack app to support carpooling to offices and back home

🧚🏻 Well, an application has also been added on our Slack Worthy leprechaunthat appreciates everyone who cares about making our offices look nice. This is what the leprechaun calls out to us now: “Look, both dishwashers are full! I'd better clean one up right now. And it's done, I finally have somewhere to put my spoon from the flower snack. “

Would you like to join us?Currently we are looking for A few colleagues right away. Get in touch with our Simona and have fun about the possibilities of cooperation: simona.slivova@futured.app & +420 735 040 126