Vision Podcast #21 – Hardware-Independent Machine Vision w/ Heiko Eisele

MVPro Media – The Vision Podcast #21

Guest – Heiko Eisele, President of MVTec U.S.

Heiko Eisele, President of MVTec’s U.S. subsidiary, outlines why hardware independence remains a key principle in industrial machine vision, how software choices shape development speed and long-term system costs, why open source involves real trade-offs in production environments, and why data labeling and management are critical to making AI work in production.

“If I would to summarize it in one word, it would be hardware independence. You have to make sure that you do a good job gathering good data. At the same time, it is really the software that understands the data and makes actionable decisions or creates value for an organization.”

On this page:

• Podcast player
• Guest information
• Useful links
• Episode chapters
• Episode transcript

Listen to the Episode:

---

About our Guest:

---

Useful Links:

• Heiko Eisele LinkedIn profile
• MVTec: mvtec.com | LinkedIn page

---

Episode Chapters

Click onto the chapters to access the relevant sections of the transcript below.

1 — From Physics to Machine Vision

Heiko Eisele reflects on his journey from theoretical physics into industrial machine vision, including his early work in X-ray CT imaging at Bosch. This chapter highlights how foundational physics, programming, and real-world constraints shaped his approach to solving imaging problems.

2 — Why Tools Matter in Vision Development

Drawing on his experience moving from academia to industry, Heiko explains the importance of having the right software tools. He discusses the trade-off between building algorithms from scratch and using established platforms to accelerate development and meet production deadlines.

3 — The Case for Hardware-Independent Vision

At the core of MVTec’s philosophy is the idea that software—not hardware—creates value in machine vision systems. Heiko outlines why separating software from hardware enables greater flexibility, scalability, and long-term cost efficiency.

4 — Open Source vs Commercial Software: The Real Cost

Open-source tools have broadened access to machine vision, but Heiko challenges the idea that they are always the best choice for industrial deployment. This chapter explores total cost of ownership, development time, and the realities of maintaining production systems.

5 — AI, Usability, and the Future of Vision Software

Looking ahead, Heiko discusses how AI is shaping both machine vision applications and the developer experience. From continuous learning and data management to improving usability for engineers and non-experts alike, this chapter explores how vision software is evolving.

---

Episode Transcript

Heiko Eisele (MVTec)

I would say if I would summarize it in one word, it would be hardware independence. You have to make sure that you do a good job gathering good data. At the same time, it is really the software that understands the data and makes actionable decisions or creates value for an organization.

Josh Eastburn (host)

Welcome to the MVP Pro Podcast.

Today, we take an almost 180-degree turn from our last episode in which we explored a design paradigm that focused on pushing more and more computing and analytics into the hardware at the edge of the network. Heiko Eisele, president of MVTec’s US subsidiary in Boston, joins me today to talk about hardware-independent system design. Heiko oversees sales, technical support, business development, and a broad partner network across North America. He brings more than 25 years of experience in machine vision, from PhD work at Robert Bosch GmbH, application engineering at MVTec, and now leading the US organization. MVTec is a leading designer of software for machine vision used in a wide range of industries such as semiconductor and electronics manufacturing, battery production, agriculture and food, as well as logistics. They enable applications like surface inspection, optical quality control, robot guidance, identification, measurement classification, and many more. By providing modern technologies such as 3D vision and deep learning, MVTec also enables new automation solutions for the industrial internet of things or Industry 4.0. MVTec is represented in more than 35 countries worldwide with locations in Germany, the USA, France, Benelux, Spain, China, Taiwan, and Korea, as well as an established network of international distributors.

Josh Eastburn (host)

In this conversation, we discuss why hardware independence matters, what the real cost of free open source tools is, how AI is changing the developer experience, and where MVTec sees machine vision heading next. With that, let’s hear the other side from Heiko Eisele.

Josh Eastburn (host)

So I see PhD in physics in your background, and I have to ask what that experience was like for you. I’ve heard that it could be grueling, of course, right? So what was your experience like?

Heiko Eisele (MVTec)

I would say maybe grueling is a good way to phrase it. For sure, I went into this realizing I didn’t pick the easiest subject, but honestly, I would say the first 2 years or so were probably the hardest, right? You come out of high school and there’s a lot of abstract math among other things that are thrown at you. Theoretical physics was tough initially. So yeah, it certainly wasn’t easy. I mean, a lot of people during those 3 years realized maybe this is not the right thing for them. But on the other hand, once you got through, I really started enjoying it. And what always attracted me about physics personally is the interdisciplinary nature of it, if you will. Physics is the basis of so many other sciences, whether it’s chemistry or biology. So that being said, I always was interested in physics and really it also led me into imaging. I mean, imaging is a physical process, right? Optical imaging, X-ray imaging. And I also got exposed during my studies to other areas of research, such as medical research, cancer research, where you apply imaging methods to do screening or whatever analysis. And I really found this interesting.

Heiko Eisele (MVTec)

Last but not least, it was also a good excuse to get into programming and software development. I always realized, you know, being a physicist, it’s tough to find a job. You’re often perceived to be as very theoretical minded unless you can bring practical skills to the table. For that reason, it was always important for me to gain those skills. So I developed an interest in programming and software as well during my studies.

Josh Eastburn (host)

So you ended up doing your focus on X-ray imaging analysis, is that right?

Heiko Eisele (MVTec)

Exactly. So this came later on. And so I was working for Bosch for 3 years and it was a program that was designed for PhD students to work on a longer-term research project that has relevance for the company. So I was actually employed by Bosch. At the same time, we collaborated with the university, University of Heidelberg in this case in Germany, which allowed me to also publish some of my work that I did for Bosch, which of course is required if you want to earn a PhD. So for me personally, this was a good segue of sort of coming from academics and making that step into industry, right? I mean, I still had the luxury of working on a longer-term project without the day-to-day pressure that some of my colleagues at Bosch were facing who had to get vision projects up and running in production. So it was a little bit less pressure. At the same time, I got exposure and saw firsthand about how things are done, how projects are managed, how machine vision projects are managed and executed in industry. And yes, I was working on X-ray imaging. It was actually X-ray CT imaging in particular.

Heiko Eisele (MVTec)

So with a technique, in essence, you’re creating a volumetric image of a part, and Bosch wanted to do that to look for internal defects. I mean, all kinds of parts, for example, castings, right? Fuel pumps and whatnot. And that process generated a lot of data. And instead of a 2D image, you can picture it like having a shoebox full of data, right? Where you have, where pixels become voxels. And then my job was to design algorithms that automatically analyze these datasets and look, for example, for defects or features that are relevant for people in production.

Josh Eastburn (host)

It’s fantastic that to have that opportunity early on in your career, that seems like a really smart program. That for Bosch to be running also internally. I wonder how does that compare to the work that you later did when you became an applications engineer at MVTec?

Heiko Eisele (MVTec)

I’d say it was similar. Later at MVTec, mainly working on visual imaging applications versus X-ray, although we also did have some customers who worked on X-ray applications. At the same time, a lot of the basic tools and methods I learned are relevant, of course, for 2D imaging as well. And even while I was still at Bosch, I realized how important it is for people to have a software package that allows them to solve applications quickly. 3D imaging or CT imaging, dealing with voxels instead of pixels, it was still kind of a niche at the time. It still is to some extent today. And for that reason, really, I did not have as much software tools available to do my work as my colleagues had. And my colleagues actually started adopting the HALCON Vision software of MVTec at that time. And I was always a little bit jealous, I have to say, when I saw how quickly they got projects done and how much having the right tools that allow you to develop quickly and efficiently matter. And I didn’t quite have that luxury. I was working with a more low-level library at that time.

Heiko Eisele (MVTec)

And even a lot of algorithms that I needed to do the analysis on these X-ray images or X-ray CT images were missing. So I had to sort of code a lot from scratch. It was a good learning experience. At the same time, I realized how much time you can spend on writing basic algorithms yourself instead of focusing really on the application you’re trying to solve.

Josh Eastburn (host)

A good experience, but maybe not one you’d want to do again, right? Not if you have a better alternative.

Heiko Eisele (MVTec)

Sure, not necessarily, especially not if you have to get projects out the door, right? And you’re facing deadlines with production launches and whatnot. I mean, you don’t want to be in that situation.

Josh Eastburn (host)

Yeah, we can all relate to that. So between your studies and working with MVTec, it seems like you’ve been back and forth between the United States and Germany, and somehow you ended ended up living in Massachusetts where you now lead from the Boston office. How did that happen? How did that come together?

Heiko Eisele (MVTec)

It was actually more coincidence. I mean, during my studies, I was in California, in Long Beach, California for 3 years. Then I went back to Germany. Good choice. Yes. Yes. Nice place to live for sure. And then I went back to Germany to work for Bosch and then I joined MVTec. I was working in Munich for 3 years as an application engineer, and then MVTec decided to open up an office in North America to be closer to the market, closer to the customers. And how did I end up in Massachusetts? I would say it was more coincidence. I mean, we were thinking about different locations in the country. We were considering where the majority of our customers are, but also what was important was the collaboration with headquarters, of course, right? And if you’re at the West Coast, you have a 9-hour time difference versus a 6-hour time difference at the East Coast.

Josh Eastburn (host)

And so I do think of MVTec as being a German company. So I guess I shouldn’t be surprised, but I was a little surprised to hear that there’s a Boston office. Tell me a little bit about MVTec’s presence in the US market right now.

Heiko Eisele (MVTec)

I mean, we’re perceived as a German company, which is good in a way, because German products generally have a good reputation. At the same time, of course, we want to be clear with the customer and we want the customer to understand that we’re here to support them locally, right? That they can call us in their own time zone, that we have engineering support, that we’re here to help them be successful with our products and get their projects out the door, right? So this is really, really important to us. So we are a small team of 8 people now here in the Boston office, including myself. And like I said, our job is to support customers with their projects, to provide trainings, and of course also like from MVTec’s point of view, to develop the market and also provide feedback from the market, North American market to headquarters so they understand what might be some of the subtle nuances that are different here than in other markets.

Josh Eastburn (host)

Mm-hmm. And so related to that, obviously each company has its own philosophy, right? Particularly with an engineering company, there’s a design philosophy. That is also part of what that company brings to market. And I’m excited to have MVTec on the show because I feel like you’re known for having a distinct philosophy, right? Compared to some of the other suppliers that are out there. So maybe we can just start this off by giving you the opportunity to explain what MVTec’s design philosophy is.

Heiko Eisele (MVTec)

I would say if I would summarize it in one word, it would be hardware independence. I mean, our belief was always that the key to solving an application is the software, right? Of course, about how you gather the data and you have to make sure that you do a good job gathering good data. If you have good data, in the end, you get good results. And if you have bad data, you will get bad results. At the same time, it is really the software that understands the data and makes actionable decisions or creates data that have value for an organization based on the original sensor data. And this always has been a focus of MVTec. And this is where the founder said, well, this is where we could basically create this library. Which later became HALCON, that allows people in academic research, for example, to get to develop applications for imaging quicker, right? By providing them tools to do that. And of course, that was sort of before MVTec was founded. And later on in the late ’90s, they decided there are customers in industry who have the same needs, right? There are plenty of applications, plenty of opportunities for machine vision to bring value to industrial customers.

Heiko Eisele (MVTec)

At the same time, time is always a factor, right? So we want to provide them a platform that allows them to develop applications quickly. And as I mentioned, basic philosophy was always hardware independence, focus on the algorithms, and it should be up to the customer to decide what hardware is best for the application. And requirements could be quite different depending what the customer wants to do. I mean, sometimes they want to inspect continuous services that are moving at a very high speed where they have to resolve very fine defects. So the challenge you have here is data throughput, right? You have to process pixels at a very high rate to keep up with the line speed. But on the other side, in some applications, power consumption might be more important, right? If you want to port an application on an embedded device, for example. So we at MVTec always wanted to make sure that our algorithms can run on any platform and the customer can basically use them and design the vision system that best suits its needs.

Josh Eastburn (host)

So I imagine a lot of people might hear this and think, okay, but there are so many new smart cameras coming out that have more and more algorithms being pushed into the hardware, right? It’s running right there on the line. I want to have the latest and greatest. Why wouldn’t I go with that approach, right? Why take the software-first approach?

Heiko Eisele (MVTec)

Well, you said something interesting. You mentioned that the algorithms that are embedded on the smart camera or run on the smart camera. So that to us actually proves the point, right, that even hardware vendors for them, software is the key and the software is what gives the smart camera, let’s say, turns a dumb camera, if you will, into a smart camera, right? And many of those smart camera vendors, manufacturers, or vision sensor manufacturers actually are our customers, right? They’re good at designing hardware and we are good at making software and why not work together? And obviously these smart sensors, they have their place, no questions, right? I mean, you get everything in one device that you can deploy quickly. But at the same time, it’s the software that in essence defines what that smart sensor can do, whether it’s smart sensor or maybe on a higher level, a trunky vision system.

Josh Eastburn (host)

And on the other hand, I’m sure there are people who are coming from the open source background, right? And thanks to packages like OpenCV in the last 5 to 10 years, right? Really growing in popularity. I think we’re seeing that’s part of what contributes to this growing interest in machine vision. Why not take advantage of those tools when, let’s say, moving into commercial applications?

Heiko Eisele (MVTec)

Say packages like OpenCV are certainly helping the adoption of machine vision and they bring it to a wider audience, right? You have a package you can use without paying licensing fees. So it really gives lots of people the opportunity to just try things out. And in a way, that’s good for us as well. Why would you not just do everything with OpenCV? Well, I would like to turn that question around. Why would you? I mean, I’m not saying OpenCV in particular, but open source, the lower-level libraries in general. And it comes back to what I mentioned earlier. It’s the time factor, right? I mean, you have a deadline to meet. You have to get a production line up and running and you have maybe a couple of weeks or so before you even see parts or see real parts, let alone defects. So the question becomes, well, what is important, right? Saving some money on licensing fees by using open source, or is it, let’s say, the time aspect and the certainty that actually in a week or so you’re done. And I dare to say for customs industry, it’s the latter aspect that’s more important.

Heiko Eisele (MVTec)

And open source libraries, again, they really help with adoption of machine vision. At the same time, generally speaking, you pay for it in terms of time. What you save in licensing, you pay in time. And you always have to look at the total cost of ownership of the solution. And that includes the development time, getting that first version of your application out the door. Or launching in production, and then also maintaining it down the road. I mean, it rarely ever happens that you deliver a project and then you’re done. Things always change. You will have to make adaptations and being able to respond to that quickly and respond to customer needs quickly to change in the production conditions. This is really what defines the value of a software package. And I think in that respect, certainly packages like ours can have an edge over open source software.

Josh Eastburn (host)

Yeah, that’s kind of what I was thinking as you were explaining that is how do you, let’s say from the sales side, even explain the ROI? That a particular customer should expect? Is it possible to place a price on hardware independence? How do you think about that?

Heiko Eisele (MVTec)

More modular design is generally an advantage for the customer because if things change, and as I mentioned, things always change, then he only has to swap out the component that he needs to change, which can meet, for example, suppose he gets a new production line up and running and he does some sort of, you know, defect inspection services and he really initially doesn’t have a good feel for what these defects will be, how big they will be and whatnot. So he starts off with a turnkey vision system and later realizes, well, my my resolution is not good enough, or I simply have to inspect different portions of that sheet to make a decision that’s relevant for production. So what do you do then? You know, if you have a turnkey vision system and you realize, okay, I’m exhausting its capabilities, you need to replace it with a different turnkey vision system, or you need to add a second turnkey vision system to cover a wider field of view. And that quickly means you realize that you only have so much flexibility. And second, you quickly actually add to the cost of the overall solution. And the real answer in many situations is, well, why not just add another sensor?

Heiko Eisele (MVTec)

Why not, let’s say, increase the resolution of the camera, change the optics a little bit, maybe get a better image, get a better resolution, and then work with the same software to address the application. And if you sort of separate the software from the hardware, I mean, you are in a much better position to address those challenges in a more cost-effective manner.

Josh Eastburn (host)

I think that seems to follow logically though, right? If you’re focusing on the hardware, expect to be investing more and more in hardware, which is harder to change. Change than the software, which is why we love software, right? Exactly. Which then leads me to another question about your products, Halcon and Merlic, the MVTec’s flagship products. Can you talk about how those products are being kept up to date? How are you evolving those tools to respond to market demands?

Heiko Eisele (MVTec)

You already mentioned market demands are important, right? We have to listen to our customers and we have to see how do they work with our products. Are there possibly gaps or things in their workflow that we could address better to make them more productive? Sometimes there are particular features that we should support to allow them to solve certain applications. And on the other hand, of course, we’re also looking at the machine vision market in general. We’re thinking about what applications could be important in the future. In what industry do we still see an underutilization of machine vision? Do we have the right tools to address these applications? What are these applications? And based on that, we also make long-term decisions. And maybe to just give one example, I mean, back in the 2000s, we already invested heavily in expanding our toolbox into 3D vision and creating tools that allowed you to do, for example, 3D position recognition with just one camera or later on also with 3D imaging sensors. And this was at a time where really applications like bin picking were still considered holy grail, so to speak, or even unsolvable, right? But we realized at some point this will become important.

Heiko Eisele (MVTec)

At some point people will move in the direction, technology will be ready at some point also in terms of sensors. In robotics, and we really wanted to be there early on. And that means often on one side, of course, we’re responding or we’re trying to respond as best we can to customer needs in the short term. But at the same time, we are also looking at the long-term market trends and we’re thinking about what areas could be relevant for machine vision in the future.

Josh Eastburn (host)

Dare I ask where the names HALCON and MERLIC came from? I have always wanted to know. Slightly off topic.

Heiko Eisele (MVTec)

Of course, HALCON is a lot easier to explain than MERLIC. So HALCON comes from the Spanish word for falcon, halcón, I believe it’s the correct pronunciation.

Josh Eastburn (host)

Okay.

Heiko Eisele (MVTec)

Yeah, and I see you’re nodding your head already. It makes sense, right? A falcon can see really well. And of course, HALCON can resolve fine defects, for example, very well, or measure subpixel resolution. And so it was kind of a good fit in a way. Actually, the falcon was also the mascot of HALCON early on. And besides that, it is a name that’s easy to remember. It sounds good in English. And so we adopted the name HALCON. MERLIC gets a little bit trickier. I mean, number one, MERLIC came to the market a lot later, so that means navigating issues like competitors, trademarks, and whatnot became a little bit more complex. And the name Merlic actually came out of an internal codename for that project at that time where MVTec decided, well, Halcon is a great library. At the same time, there are customers and users of machine vision out there that are not programmers. So this is why we came out with Merlic. And the name Merlic came, as I mentioned, from an internal codename for that project prior to release. And then of course, as we got closer to launch, we said, what do we do with that?

Heiko Eisele (MVTec)

Now we have to have a real name that we also put in the website and we put in our literature and whatnot. And there you have to start, well, are there any trademarks that you possibly violate, or does it sound similar to an insult in some language? I mean, we’re selling worldwide, right? Does it sound similar to something that you do not want to be associated with? And then in the end, somehow that became Merlic. So it’s hard to really associate something with it, but in the end we felt, okay, it’s still a name that’s short, that’s reasonably easy to remember.

Josh Eastburn (host)

That’s just the reality of international business, I think, huh?

Heiko Eisele (MVTec)

Yes, yes.

Josh Eastburn (host)

So getting back to the topic at hand and looking, I think, a little bit into the future now, right? So this is, we’ve talked about a little bit about how these products have evolved. Where do you see them going in the next few years? What’s on the roadmap that people might look forward to?

Heiko Eisele (MVTec)

Maybe you can start with Falcon. As I mentioned, we always look at whether there might be gaps in our customers’ workflow that we need to address. And for a library like Halcon, this could mean, for example, we need a new operator, right? It also could mean we gotta make some adaptations in some language structures. And Halcon comes, of course, as a library, but also it comes with a development environment that customers use to quickly deploy applications. And here workflows become more and more important and productivity. And here we are actually right now in the process of revamping our edge develop IDE, Integrated Development Environment, because of what we call edge develop Evo. And we’re modernizing that environment in a way we’re supporting new kind of language structures and just making various improvements that allow customers to be more productive in the end. Everyone is talking about AI and we have to realize that AI is also becoming more and more important as a tool for software developers, right? For example, to quickly search the information that they need. The Halcon documentation is probably about, I have no idea honestly about, but maybe about 10,000 pages long and finding the right piece of information that you really need in a given situation is becoming more and more of a challenge.

Heiko Eisele (MVTec)

This is where AI can help and we’re looking into that. And of course, AI will also play an important role in coding and maybe to some extent replace coding, but at the very least, it will help people to come up with the initial prototype. And we’re also looking at integrating these tools into our IDE.

Josh Eastburn (host)

Yeah, I guess that gets back to the theme of productivity that you mentioned earlier, right? Helping engineers deliver on time.

Heiko Eisele (MVTec)

Yeah, exactly. And then as far as the libraries is concerned, there are always things to improve in technologies that we have supported for many, many years, including code reading, for example, optical character recognition, or 2D pattern matching tool, for example, we continually improve it. Then of course, all AI, I mean, a lot of investment goes into AI and we are not really trying to say we’re an AI company, really. We want to make AI usable to our customers and we want to think about for which applications is useful. And we approach it from that angle. We recently came out with what we call continuous learning. So one of the challenges that you have with AI application is that number one, you need a lot of training data. And when things change a little bit, you need to add new training data and your training dataset becomes bigger and bigger and bigger. And to train it, you need a lot of GPU horsepower, which you later don’t necessarily have in the production system. So continual learning allows you to quickly retrain a model just by basically adding to your existing dataset and doing what we call incremental training.

Heiko Eisele (MVTec)

So you have the model, you add a little bit of data, and then you adapt the model instead of training it from scratch. And we’ve introduced it in the recent release last November and for image classification. And going forward, we’re certainly looking into other areas of deep learning as well, where customers can take advantage of that. Besides that, as I mentioned, I mean, we are always looking at how can AI effectively be used to solve certain tasks. I think in robotics there is still potential. 3D recognition, position recognition of objects, 1D, 2D barcode reading. Those areas where AI can provide improvements and we’re looking into that, of course.

Josh Eastburn (host)

How about Merlic? What do you see coming in the future for that product?

Heiko Eisele (MVTec)

Well, for Merlic, we have to keep in mind that our target audience is a little bit different. So it’s a very powerful, very flexible platform with the latest and greatest image processing technology. With Merlic, you know, we have to go beyond that and we have to think about, okay, how can we really provide that technology to users who are not expert in machine vision, who are not programmers, right? So our goal is for MERLIC is to bring the quality of HALCON as far as the image processing and wrap it into tools that are easy to parameters, easy for people to use with less knowledge of machine vision, let’s say production engineers maybe. And from that perspective, so of course the ease of use is critical for MERLIC and also with MERLIC, we are going beyond just providing a library that users integrate into their existing software application. We actually want to provide an all-in-one solution that people can deploy in production, kind of starting from scratch, if you will. And with that being said, of course, we have to provide much more than just a library. We have to provide a user interface or the ability to build a user interface.

Heiko Eisele (MVTec)

We have to provide communication interfaces. We have to think about tools that are relevant for monitoring things in production and all of that. So I would say usability, always important for us, right? But in in the context of Merlic, we have to think about usability different than when we talk about Halcon. For Halcon, it’s usability for the programmer, for the software developer, and making them productive as possible. For Merlic, we have to think about the production engineer and what does he need, you know, beyond just, let’s say, the core vision. And that sort of guides our roadmap for Merlic as well. So our goal is always to kind of bring the technology from Halcon into Merlic, yet at the same time, we have to think about, you know, how do we make this accessible for non-experts? And I mean, going forward, we will continuously bring new functionality that we have developed for Halcon into Merlic. Let’s say on the deep learning front, for example, there’s quite an extensive toolset available in Merlic already. And of course, other tools as well. And as I mentioned, beyond that, issues like communication interfaces are very important. We want to make sure we support all the communication protocols that customers are using in a production environment.

Heiko Eisele (MVTec)

And there are certain aspects of the user interface that are important, such as For example, having an image history tool where you can change parameters in the application and see how those affected, say, the previous 10, 20, 50 images. So all these issues are important. And like I said, with Merlic, we have to take a different perspective on what usability means and what this means for the particular target user for Merlic.

Josh Eastburn (host)

So you’d mentioned some things on your roadmap for AI. What specifically are you doing around those concepts, machine learning, deep learning? How are you bringing those to market for people?

Heiko Eisele (MVTec)

Important for AI applications is labeling, and labeling data in a way replaces writing code for classical applications, right? And then the question becomes, how can customers effectively and efficiently label data? How can they collaborate? I mean, data labeling is often done by many, many people in the factory. And to support that, we have a third product besides Hagen and Merlic. We have the deep learning tool, which supports exactly that. So it allows customers to label data. And it allows them to train models that they then can deploy in Halcon or in Merlic.

Josh Eastburn (host)

I think that’s pretty interesting because I’m sure we’re all still educating ourselves on how this technology works. And one of the things that we have tried to bring light to on the podcast specifically is the reality of working with AI, so-called AI technologies, right? And I think the process of specifically what you’re talking about of labeling data is probably something that’s underestimated by a lot of users. Is that what you see? Is that why this exists? Let’s say.

Heiko Eisele (MVTec)

Yes, in a way. Exactly. And it’s the labeling, but also the gathering and the management of data in the first place. And often labeling data is not enough. I mean, you always have to think about augmentation, but not, for example, you have to think about, okay, data distribution. Did I cover all use cases and so on? So where do the data come from? Let’s say you have several items or several stations of a given application deployed. You want to maybe keep track of where the data come from. And which system produces outliers. So just the managing of data is key, really.

Josh Eastburn (host)

This tool, the deep learning tool, runs, let’s say, in parallel with either Halcon or Merlic just to help manage the data that’s being fed into those systems. Is that right?

Heiko Eisele (MVTec)

Yes, exactly. You manage the data, you train your model, and then you deploy the model in Halcon or Merlic.

Josh Eastburn (host)

Fantastic. It sounds like you have a really clear idea of what each tool is for, right? Who it’s for, right? And tailoring the roadmap to meet those needs continuously. Very cool. If I’m a listener, I’m hearing all of this and thinking, okay, I want to give this a try. Is there like a trial license or something like that that people have access to, to make it work?

Heiko Eisele (MVTec)

Yes, of course. As I mentioned, customers want to try the software before they commit to it and make sure they actually can solve their application. So yes, we do provide free trial licenses. Not only that, as I mentioned, we also offer the help of our application engineers during those trials to make sure the customer really doesn’t waste time, but really can come to a solution quickly and has a good base to make a decision. Go to our website, download them, but we always encourage our customers, don’t just, you know, try it yourself, contact us and let us help guide you a little bit. So when we talk to customers initially, we really try not to be too product-centric. We first try to understand the customer’s workflow. I mean, what are they looking for? Are they looking, let’s say, to standardize on a machine vision platform across the factory where they have to address many different applications? Are they just looking for one specific solution for now? Are they trying to launch a product that uses machine vision? And based on that, we start the conversation and see what’s the best fit. It really also comes down to what resources does the customer have?

Heiko Eisele (MVTec)

Does he have programmers who can adopt a tool like Halcon, or does he maybe need the support of a system integrator? So we try and guide them on that, on that basis. And of course, also then as the customer really tries that product, I mean, customers generally, they don’t wanna buy a toolbox, right? They wanna buy a solution. Or at least if they buy a toolbox, they want to have confidence that it actually does what they need it to do. And this is exactly why we have application engineers here that even help customers make the right choice here before they invest money in the software, before they invest a ton of time in adopting it. Yeah.

Josh Eastburn (host)

So last year you had a great set of demos running at Automate. Are we going to see you at the show again this year?

Heiko Eisele (MVTec)

Yes. Yes, of course. The Automate, honestly, has come a long way and A3 has a great job, I think, doing a great job promoting that show and being an ambassador of automation in general in North America. And it’s impressive to see that show grow year over year. And of course we’ll be there. It’s our flagship show here in North America at this point. We’ll have demos again, of course. And each year we are at the Automate, actually for the majority, we’re talking to customers we’ve never spoken to, which is really encouraging as far as what it means for automation of America, right? I mean, there are more and more customers are really looking to automate and looking to adopt vision. And some demos which we had last year, you’ll see them again, just because they are so relevant for so many people, like a code reading demo, for example. At the same time, of course, also wanna show something new. So expect that we will be showing some basic technologies like code reading, some maybe more sexy stuff, if you will, like AI. We also will bring something new to the table, which I think we’ll announce in the months to come.

Josh Eastburn (host)

Fantastic. And so you mentioned the website mvtec.com. Are you active on social, YouTube, anything other places that people can find you?

Heiko Eisele (MVTec)

Yes, I’m actually not very active on social media have to say

Josh Eastburn (host)

Personally, yeah.

Heiko Eisele (MVTec)

I always tell myself I spend so much time in front of a computer every day. I have no desire kind of to do this in the evening as well. But if you want to connect to me on LinkedIn, feel free to do that. And at the same time, yes, we have a website. You can reach out to us here, to our team here in North America. And there are plenty of resources also on our website from success stories to white papers, educational material, and even the MVTec Academy, which is an online learning platform to get acquainted with our tools. So many options out there.

Josh Eastburn (host)

Yes, I can completely understand not wanting to spend all your time on social media. I think that’s probably healthy, but how about the company itself, MVTec? Are you very active on LinkedIn or other channels?

Heiko Eisele (MVTec)

Yes, MVTec actually is very active on LinkedIn, so we have plenty of followers and we constantly post content, whether it’s about events or whether it’s about, let’s say, you know, good customer stories. So encourage you to follow the company to learn more and stay up to date.

Josh Eastburn (host)

Well, thank you very much for your time this morning. Is there anything I neglected to mention?

Heiko Eisele (MVTec)

No, I think we covered a lot of ground. Appreciate the time and I hope that, you know, listeners will find it interesting and yeah.

Josh Eastburn (host)

Yeah, no doubt. Thank you very much.

Josh Eastburn (host)

That was Heiko Eisele, president of MVTec North America. Thanks also to Markus Setzer at MVTec for all his work in the background to make this interview happen. If you’re curious about how to put these ideas into practice, don’t stop here. Connect with Heiko on LinkedIn, visit MVTec Software’s website. You’ll find deep dive resources on HALCON, MERLIC, and the new deep learning tool, plus success stories and training material to help you explore the potential of hardware independent system design. If you are a vision professional or an integrator with a unique story, something that moves beyond the hype and into the day-to-day reality on the plant floor, I’d love to hear from you as well. Reach out to me at josh.eastburn@mvpromedia.com

Josh Eastburn (host)

Mvpromedia.com or find me on LinkedIn. For MVProMedia, I’m Josh Eastburn. Be well.