Founder and director Dennis Schipper: “We have been making smart use of technology, materials and data from the very beginning. This was mainly done through separate projects at first. We are implementing AI in all our branches and markets as part of our current program. Wherever there is data, AI is of value.”
“Wherever there is data,AI is of value.”
David Rijlaarsdam, Business Unit Manager Artificial Intelligence is responsible for this program. “On the one hand, the program allows us to share what we are doing with the outside world so that parties who need AI for certain applications know where to find us. On the other hand, it is meant to boost the company: where do we want to be in terms of projects in five years, which people should we hire, what tools should we develop? It is an integral program within Demcon.”
Foto David Rijlaarsdam
Robotic arm
Rijlaarsdam thinks that the robotic arm is a great project example. The arm finds its way into water pipes to determine their quality. “The image recognition used for path planning now takes place via a deep neural network. The robot recognizes what kind of pipe element is ahead based on this: a curve, a T-junction or something else. The robotic arm is completely disconnected from the outside world, so it needs to be able to decide where to go on its own.”
The robotic arm is fed with synthetic data to teach him this. “We taught the robot what a pipe looks like by supplying it with many fake pictures. This is cheaper and faster than building an entire pipe system. It is also much more effective than learning in an existing system, where it can take a long time for the robot to encounter all possible situations.”
Photo robotic arm
Data synthesis is also used in the agri-food sector. AI can be used to create unmanned strawberry picking systems, for example. These systems can recognize strawberries, determine their exact location and whether they are ripe to be picked – all of which they learn from synthetic data. Demcon has its own animation studio, Nymus3D, which specializes in mimicking the complexity of nature, among other things.
Video: synthetic training data (agri)
Video: Nymus3D demo reel
Eggs and coins
Another example is the classification of eggs. An automated process assesses whether 40,000 eggs per hour fall in category A or B. Rijlaarsdam: “The great thing about this system is that it can be easily expanded by the customer, even if they have no knowledge of machine learning. We supply software with the system that allows you to easily add a new category of eggs, for example, after which the machine learns to apply it by itself.”
Rijlaarsdam also wants to mention the example of the coin machine. This is the request that Demcon received: determine whether 40,000,000 coins (no, this is not a typo) are damaged or not within a month. Without any data on when a coin is damaged, exactly, and with a reliability of 99.9999% (again, not a typo). Rijlaarsdam: “We only had a model of what the perfect coin looked like, and we made a sorting machine with a deep learning network which compared the coins to this model. We succeeded and managed to assess and sort all the coins within a month; this amounts to about 15 coins per second.”
Photo coin machine
Getting knowledge to SMEs
Dennis Schipper is equally enthusiastic about the examples. However, he also realizes that such AI applications are not yet within reach for everyone. “These kinds of projects are successful because of all kinds of promising technologies. These technologies are quite accessible for a company like Demcon, but it is difficult for smaller companies to fathom all the possibilities.” He adds that the AI hub East Netherlands plays an important role in this. “Having a partnership from which you can valorize and spread knowledge is very helpful. You can also go to them with questions or projects.”
“You can move towards application by connecting parties that deal with AI.”
Photo Dennis Schipper
A good balance is important: “We must continue to expand our knowledge, and we must ensure that we get that knowledge to other companies, especially the SMEs. That is the most important target group, often with regional ties and a great need for a link to educational and research institutions, for example. You can move towards application by connecting these parties.
The ideal image
According to Schipper, the hub can play a role in three areas: projects, networking, and money. “The people at the hub must be able to define a project together with an SME, and a knowledge institution must be involved – such as an educational party or Demcon – and some funding must be made available. That is the ideal image. It is then up to the companies themselves to ensure that it is set up effectively.”
“The hub ensures that companies and educational institutions listen to each other more.”
“In any case, it is good that the ecosystem is developing”, he continues. “More attention for the topic creates more interest, more interest creates more students, which in turn leads to more well-trained staff. The hub can also ensure that companies and educational institutions listen to each other more.”
