Wakoon

Wakoon
Azure machine learning

Inspiration

We wanted to make our lives easier by using recent technologies as the "Internet of Me". The hardest thing for us is to get up in the morning and get into our daily routines.

This means, we want the most important informations about the day in a short way and to change the alarm clock automatically, if the schedule is changing.

But the most powerful change in our live would be an automatic selection of the clothes. That's why we wanted to make our wardrobe smart. As a team of data scientist and programmers, we were a perfect combination as this tasks needs a good User Interface and smart algorithms.

What it does

Next to looking for the schedule and rescheduling of the alarm clock, we integrated a smart outfit selector, which learns the preferences of the user. It learns, which clothes the user likes and which clothes to wear if it is raining this day and to react on the temperature. Based on that it checks, what is in the wardrobe and calculates the probability of each item. Then the whole result will be given in an android app, which the user can use in the morning.

How we built it

The android app was written in Java, while the machine learning part was completly done in azure. The wardrobe was simulated by a sql table. Next to the machine learning and data manipulation functions of azure, we coded several script using R and Python, which perfectly fit with the azure environment. As azure provides us with a web service, the connection to the app is handled by the azure API.

Challenges we ran into

We had to face many challenges, as we were not used to azure. The communication of the program was a hard problem, as we had to learn a lot about azure. In addition, we had to learn how to work with our teammates, as we had completely different backgrounds.

Accomplishments that we're proud of

We created a functional App with a clean UI that contains the planned features. We use strong machine-learning algorithms in the back-end and integrate with other web-services. By use of the cloud we are easily scalable and deployable. We really think this app would make our mornings a bit better. With the used architecture we are able to help many other people with the same problem.

What we learned

On the technical side it was our first contact with Azure and we learned a lot about that ecosystem. We use databases there and run all of our machine-leaning on the cloud, while using our own scripts where appropriate. We commnicate with all of this from directly from the Android device. As a team we merged groups of different technical backgrounds which made us think from other points of view.

What's next for Wakoon

Data-collection can be performed in a smart way. We have access to different sensors we could use to make the programm know exactly what lies in the wardrobe at a given time. With more and better data the machine-leaning component can be greatly improved. Having more data, we could use the information about the preferences of the people in the same region or same country. It is also possible to have a look on future events and the content of the wardrobe, such that the app will inform you when to do laundry ( for example, if you run short on shirts, or if it is raining for the rest of the week and there is only one sweater left, or a date is coming up and you have no appropriate clothes ).