Intro
As March was approaching, I started hearing some buzz from my network of friends that attend hackathons. They were talking about this ABInBev hackathon with huge prizes and I was honestly not very interested. I’m not much of a beer fan as much as I like wine and I tend to avoid competitions that are centered around solving problems for a specific company… it makes me feel like I am being bought out? Which I probably am. Anyway, I had nothing to do that Saturday so I ended up going.
So what is this hackathon about?
Website Description:
Anheuser-Busch beers must pass through multiple checkpoints before they can be enjoyed by consumers. After being brewed, bottled and packaged, beers are shipped to wholesalers who then have the mission of selling them to various POCs or points of contact – think your favorite bar, restaurant, grocery or liquor store. The process to get beers from the brewery into consumers hands has improved tremendously over the years but inefficiencies still remain. Help us bring innovation to POCs and deliver your favorite beers to you as quickly as possible.
Basically, they are looking for solutions to their distribution network and guaranteeing that their beer will always be stocked in a store.
Our Solution
Sajal, David and I came up with a project called shelfie. It is a beer stock monitoring system which is installed in the fridge shelf of the store. The camera will detect when the beer is out of stock and alert the beer manufacturer. It also detects the type of beer such as Stella Artois, Shock Top, Elysian, etc. Another added benefit is that the system will guarantee that the beer is located in the correct placement on the shelf and competitor beers are not located on the same shelf space.
How does it work?
The system consists of a Raspberry Pi micro computer which is installed in the store shelf and a web server to receive data from the shelfie systems for analysis. First, the Raspberry Pi captures images of the top of the beer bottles. The image is processed to find all circular looking objects in the picture which should look like the top of a beer bottle. Then the circular objects that are found are compared against a classification algorithm to determine if it is a recognized beer brand. Finally, the images of the bottle tops and location data are uploaded to the web server to be displayed on a UI showing the beer arrangement on the shelf.
Check it out
The full project is located on our devpost page : Shelfie
Here is the presentation video
Pictures
Math/Comp Sci Blog 