5 teams - 5 topics - 24 hours

The Sweet Mustard Innovation Day is a 24h quest for inspiring technological experiments, stunning design ideas and surprising business insights. The kick-off of our latest edition was given on Thursday evening, November 8th at the Sweet Mustard headquarters. The new concepts were presented 24h later during the Innovation Day Awards @Hangar K.
Are you curious about the outcomes? Read on to have a sneak peak on the work in progress.

Image1 1

Team code@school

In preparation for an interactive workshop to inspire children for technology, we experimented - amongst other things - with Blockly. Blockly is a framework from Google to present programming more simply. We will use this to control our Barbot, a cocktail-making robot.

Children can use the blocks to access the technology that will configure a cocktail of their choice, after which the cocktail will be served. Child-friendly, without alcohol of course.

Besides the Barbot, we have also prepared something with the Nao Robot (hint: the floss), 3D printing of our mascot and how we can make children discover Artificial Intelligence (AI) in a simple, but challenging way.

At the end of the day we also created an exciting programme so that the children can be optimally stimulated by technology and of course enjoy a nice mocktail. Cheers!

Image2 1

Team matinga

Meet your goal with Matinga

Internal research has shown that the average meeting has a productivity score of only 64%. Moreover, 75% of the respondents spend more than 4 hours a week in various meetings. This seemed like a good opportunity to explore different issues with meetings and tackle these problems. We mixed our ideas with some gamification and developed a card game for meetings. The purpose of the Matinga project is to help establish one or more of the following targets:

  • Have a clear meeting goal
  • Keep participants engaged
  • Follow up on action points

After defining a goal, the team members worked on ideation, prototyping and testing. In the course of the day, Matinga was born: a meeting game to meet your goals.


Team crowd game

Our goal was to create a game that is controlled by a crowd, or a small group of players. We had various fun and cool ideas to achieve this. By applying machine learning, we could for example, write software that detects arm movements of a crowd to steer a protagonist in a game, or maybe create huge joystick or controller that is controlled by multiple persons.

We decided to create a Pac-Man game, custom build a big controller to steer Pac-Man, and develop a Progressive Web Application to control the ghosts chasing Pac-Man. So we could have 4 persons working together to save Pac-Man from the many ghosts chasing him. Everybody could pick a ghost, jump in the game and try to catch Pac-Man.

We found an open source JavaScript implementation of the Pac-Man game with a license that permits us to alter the game. The game had to be adjusted to make the ghosts controllable. Each player has four movement controls (left, right, up and down) e.g. key A moves Pac-Man to the left, B moves it up, key E moves ghost 1 to the left, key F moves it up and so on. The game can start when at least two players have pressed one of their four buttons. To make the user interface more customizable we embedded the game in a React app.


Team warmste weegschaal

It began with the idea to bring people together in an entertaining way by letting them interact.
It’s not really in the nature of the average Belgian to simply go and interact with others. So we wanted to trigger people enough to get them out of their social bubble.

And BAM! The “Warmste Weegschaal” was born!

Warmste Weegschaal? What is it?
It’s a combination of both physical and digital scales that are linked with each other. They offer the possibility to give your opinion on specific topics by voting.

How? By donating money to a good cause. After you’ve donated a certain amount of money (either digitally or cash), you are given coins to place on the scale. The weight of the coins will tip the scale to the winning side or opinion. It’s that easy!

Want to see your own topic or question for others to vote on? No problem! We’ve got you covered. You can buy a topic just like you can buy coins for voting.

We have chosen to donate the benefit of the Warmste Weegschaal to Warriors Against Cancer VZW from Kortrijk.

This organisation helps people who have been through their treatment with sometimes permanent injuries -physically and / or mentally - to continue their lives. Thanks to the donors, they realize - amongst other things - “Spread Your Wings Photoshoots” and “Feel Good Days” and thus provide support in improving the quality of life of people and in their reintegration into society.


Team clean desk police

Ever been busy with something, then something else and then some more, … When you finally come back at your desk, it’s still a mess from the last time? We know the feeling! During this Innovation Day we focussed on innovating our desk. To keep itself clean(er), so we can do other (more) fun stuff!

This could be realised by flipping our desk over, sliding everything in a collecting basket. The basket keeps everything in one place and we have space to work when we get back. As soon as you’ve got some larger equipment that you forgot to mount, like a monitor for example, this could get expensive.
So we looked at a more delicate and refined solution: a robotic arm equipped with a camera perhaps? This could ensure us some work for the next couple of months, but the time was limited to 24h. So we decided to focus on the intelligence of detecting when a desk should be cleaned and what should be cleaned up.

We first took some pictures of a clean desk and a messy desk, to see what should be removed. We used OpenCV together with Jupyter Lab to prototype some algorithms. Comparing the clean desk with the messy one, already gave us some nice results. Together with some filtering methods, we detected all the ‘junk’.

Image9 1

Detecting all the different parts was already a big step, but without knowing what it is and where to put it, we got some more work to do. The next step was recognizing the parts.
We looked at feature detection to map possible items with the items on the desk. The key points matched pretty well, but there were still some issues.


Machine learning could also have been an option, but we didn’t find a good dataset to train our model. By the time we had taken additional pictures of the different tools, the day was already over. But we will continue to develop the solution. Stay tuned to learn more about it!