DeskBot

Inspiration

For this themeless hackathon we created our own theme: Productivity and Health in the Work Place. We are often tempted towards distraction while trying to be productive. The Deskbot is designed to monitor unproductive behavior and notify you. In an attempt to reduce back injury, we monitor posture and notify you when it becomes bad.

What it does

Deskbot uses a combination of software and mechatronics to notify the user when they are off task on their computer, slouching in an unhealthy way, or if their desk is especially messy.

How we built it

First we take images from the webcam and the screen and using a retrained MatLab neural network, deskbot identifies if you're being productive, slouching, or if you need to clean your desk. If one of those are true, it sends that data to the Arduino, which first warns the user with a flag, then proceeds to beep and blink if changes aren't made.

Challenges we ran into

Figuring out how to identify what was productive, and training the neural net to accurately identify it was a large challenge. On the mechatronics side, figuring out how the computer communicated with the Arduino without being able to see the actual data because the Arduino only had one serial port (used for communication) was also tricky.

Accomplishments that we're proud of

Getting the neural net to be able to identify it's parameters with high success and while looking at several different people.

What we learned

Improved our skills with MatLab neural nets and Arduino and communications between the two and how to pull an all-nighter.

What's next for DeskBot

Redesign housing using a 3D printer, create a user-friendly application to collect images for training and to use the software. Make the setup more compact.

Built With

• arduino
• matlab