BeerBot

photo @ 11:57
photo @ 12:08
photo @ 12:24

Inspiration

A never-ending supply of cold beer at KI labs to maintain employee happiness.

What it does

Monitors the beer refrigerator and uses image recognition to determine the number and coldness of the beer bottles inside. It ensures that realtime beer updates are continually tracked. An important feature is the capability of the system to proactively alert when the beer supply is critically low.

In short, it is a modern, reliable, scalable solution to an everyday problem - "do we have beer? is it cold? should we reorder?".

How we built it

We designed a solution to be non-intrusive and require minimal maintenance. This was only possible with the following hardware "modifications":

headless raspberry pi
infrared camera
calibration of camera
infrared illumination for camera quality

The underlying pipeline consists of an Image Processing Engine connected directly to a smart slackbot. The overall processing is as follows:

Door Detection: state of fridge identified (closed vs. open)
Bottle Detection: image recognition based on thresholding, segmentation and superpixels
Bottle Tracking: combine temporal detection output to identify "age" of individual bottles
Slackbot Capabilities: explicit state requests, automatic inventory update notifications (e.g. bottle tracking, low supply)