Bar-Duino

Bar-Duino Prototype
Face Recognition on web app
Drinks Page
Drink page

Inspiration

We believe strongly in 2 things 1. Alcohol 2. Efficiency, so the idea came quickly to automate the process of serving drinks in an entertaining way...

What it does

Facial recognition scans the user's face to check they are old enough to drink and remembers your face to build a profile on you. Once logged in, it directs them to order a drink, where they can choose from a massive range of drinks, with filtering depending on what ingredients we have available. In order to personalise, the app will ask the user to give feedback on their previous selection and give personalised recommendations based on all the answers you have given. Once you have chosen the drink, it contacts our brilliantly designed setup to pour a certain amount from the necessary spirits/mixers using an Arduino some servomotors and custom 3D printed parts to form our pouring/stopper mechanism. Despite all of this, our favourite part of this project is the beautiful LED with a variety of themes to choose from such as Midnight, Tiki to Golden Hour.

How we built it

We began by getting the skeleton of the web application in HTML and CSS while we developed the servomotors so they could be sent the necessary commands to pour from all our bottles. At the same time creating a CAD design for the rotatable covers. We then made the backend for the web pages in javascript, including facial recognition with face-api.js and implemented a database using firebase to store the user data. After much bug fixing, with both the physical setup and the web page, we were ready to connect all the physical components and deploy the website on firebase.

Challenges we ran into

Our mechanism for covering the opening is rotating the cover in a plane parallell to the opening on the outside. This has a high chance of leaking, since we can never get it flush to the surface as there has to be some tolerance. To solve this, we sealed the area around the hole with patroleum jelly. We also encountered some issues in the inital setup of the web page. We intended to use the python Deepface API, so began building with React and python. However Deepface didn't work when we tested it. To fix, we looked for an alternative, which we discovered in face-api.js, so we switched to html, css and javascript for the backend.

Accomplishments that we're proud of

Face recognition. Sucessfully sealing the motors. Surviving with very few hours of sleep.

What we learned

We spent a little too long trying to get Deepface to work without actually checking it exists in github anymore.

Next time we would look at a vertical bottle covering mechanism, as it would be easier to make secure without having to seal with vaseline.

What's next for Bar-Duino

We would make a better prototype with the outlined changes. Allow manager to input what drinks are in stock in each bottle, as well as changing number of bottles. Add weather data in to add to increase specialisation of suggestions.