Inspiration
Our motivation to join the challenge began with the description of the project, which searched for data driven decisions using data science tools and analytics to solve real world problems. As data science students, we are excited to start applying as soon as possible what we learn inside class environment into real-world solutions. After learning about Gategroup, their mission and the challenges they face, we were drawn by the need of efficiently detecting the expiration dates in products, and we were excited to tackle this problem using the tools and knowledge we've been building while attending Gategroup's needs.
What it does
Our project digitizes the control of expiration dates in airline catering. Each food tray and trolley receives a unique barcode, allowing employees to quickly scan and register the products, their batch numbers, and expiration dates. This barcodes allow tracking the nearest expiration dates efficiently, reducing the need of constant manual verifications and lessening time spent in these manual checkups. The system automatically updates when items are replaced or removed and sends an alert when trays in a trolley must be checked for possible products with nearing expiration date. Our system reads printed expiration dates directly from packaging through Computer Vision, eliminating the need for manual checks and reducing human error. There's no need to examine all expiration dates before every single flight, this is just necessary when the system alerts us, so it is certain that there might be a product that should be removed. And specific products in specific trays and trolleys are pointed out by the system, limiting the amount of products that should be verified, enormously reducing the time and effort used in this process.
How we built it
To build it, we thought about a solution that is simple, effective and not overly complicated. We asked ourselves, how can we make this system efficient, without compromising the needs of Gategroup? We came up with the idea of delaying actual verifications until these were completely needed, as well as limiting the actual range of products that are checked. The manual process that Gategroup currently has is extremely ineffective because there's no track of any expiration dates over the trolleys and trays they are offering. They could be checking for expiration dates in products that were recently given, and which expiration dates is not even remotely near. They could be missing some expiration dates when checking so many products. They could be checking an entire trolley when maybe just one tray needs to be checked. There's no regulation, and we detected this necessity, and exploited the current structure of Gategroup.
Challenges we ran into
We faced the challenge of designing a solution that fits within Gategroup’s existing process (optimizing, not replacing). Also, of course we had some normal problems with implementation at different stages Integrating vision-based date detection across different packaging formats and lighting conditions required significant fine-tuning.
Model assumptions
- When a product in a tray is filled, all products have the same expiration date.
- The system alerts when the expiration date is in the following 5 days.
- New products that are filled in existing trays must be filled below older products. This is essential for verification warnings that are more spaced and are not so continuous.
- We do not assume that a product with certain expiration date is finished until a verification is made.
Accomplishments that we're proud of
We're really proud of all the things we achieved as a team, from hour one at taking notes, proposing different ideas, discussing them, and viewing pros and cons, to dividing tasks so we could work in parallel and boost our productivity. We are really proud of getting out of our comfort zone which is data science and let us dive into software development domains. But most of all, we are really proud of each other, because even though it was only a weekend, we've grown up a lot by learning new things and who we have became.
What we learned
We all learnt so much things about teamwork methodology, such as dividing tasks into sections like backlog/doing/done, and how planning is one of the most crucial parts of a project and delivering a good result.
Log in or sign up for Devpost to join the conversation.