At Automate 2026, artificial intelligence dominated conversations across the exhibition floor. But while many companies were eager to position themselves as AI-first, MVTec’s message was refreshingly different.
The company, celebrating its 30th anniversary this year, believes the future of industrial machine vision isn’t about replacing proven technologies with AI, it’s about combining them.
Speaking with MVPRO Media at Automate 2026, Heiko Eisele, President at MVTec LLC, explained that AI is becoming an increasingly important part of machine vision, but it should be viewed as another tool in the engineer’s toolbox rather than the complete solution.
“We’re not an AI company. We always see where we can leverage AI to add value to the customer.”
That philosophy runs through every aspect of MVTec’s software portfolio. Rather than building systems that rely solely on artificial intelligence, the company focuses on blending deep learning with classical machine vision algorithms to deliver the reliability, speed and precision demanded by modern manufacturing.
Why AI Alone Isn’t Enough
Deep learning has transformed applications such as quality inspection and defect detection, making it possible to solve problems that were previously extremely difficult.
However, Heiko believes AI has limits. “In manufacturing,” he explained, “AI may get you ninety percent there, but for the last ten, or even the last one percent, classical vision is still essential.” It’s an important distinction.
Manufacturers don’t simply need software that works most of the time. They need systems capable of delivering repeatable, highly accurate results every production cycle. That often requires traditional vision techniques for tasks such as alignment, pre-processing and precision measurement, while AI provides robustness when dealing with variation.
“Really, we barely see AI-only solutions. AI is one of many tools customers use to solve an application.”
Making AI Practical
One of the demonstrations showcased how AI can inspect components using only images of good parts.
Instead of requiring thousands of examples of defective products, the software learns what a normal product looks like before highlighting anything that appears unusual. This approach is particularly valuable during production ramp-up when manufacturers may not yet know what defects will eventually occur.
The system even understands context.
A small scratch on a cosmetic surface may be acceptable, while an identical mark in another location could indicate a critical fault capable of causing product failure. Rather than treating every anomaly equally, the software learns which defects matter based on where they appear.
Designed for Non-Programmers
Another key theme throughout the discussion was accessibility. Machine vision has traditionally required specialist engineers and significant programming expertise.
MVTec wants to change that.
According to Eisele, users simply need to understand what constitutes a good part and a bad part. The software, for example, MVTec MERLIC then guides them through the rest of the process. “We’re also targeting the non-programmer. People who are not vision experts want simple tools they can connect together in an intuitive way. They want to be up and running quickly.”
Reducing engineering complexity doesn’t just save time, it helps manufacturers deploy automation faster and respond more quickly to changing production requirements.
Speed Matters
Industrial vision systems don’t have the luxury of slowing down production. Inspection must happen within milliseconds. For many applications, MVTec’s software can process inspections in around 70 milliseconds, with additional optimisation available for even faster production environments. Other demonstrations highlighted equally demanding challenges.
Barcode and code reading may not sound exciting, but reading damaged, scratched or distorted codes in only a few milliseconds, often on embedded hardware, is far from straightforward. Rather than forcing customers to replace their hardware, MVTec’s software-only approach allows manufacturers to upgrade existing systems while maintaining compatibility with Windows, Linux, x86 and ARM platforms.
That flexibility, of a software like MVTec HALCON, reduces costs while extending the life of installed equipment.
Bringing AI Into 3D Vision
Perhaps one of the most impressive demonstrations involved MVTec’s Deep 3D Matching technology. Using CAD models instead of extensive image datasets, the software generates synthetic training data to recognise objects for bin-picking applications. Multiple cameras identify objects from different viewpoints before classical vision techniques refine the final position to deliver the accuracy required for robotic handling.
It’s another example of AI providing robustness while traditional vision delivers precision.
Looking Ahead
As AI continues to reshape industrial automation, MVTec sees its role as making sophisticated vision systems easier to deploy – also with the support of AI. Future developments include further simplifying deployment for non-programmers while also integrating AI-assisted coding tools into its software environment to accelerate development.
Ultimately, the company’s goal remains unchanged after three decades.
“You have to make the customer productive. You have to keep the engineering time to get the vision system onto the production floor as soon as possible.”
In an industry increasingly captivated by AI headlines, MVTec’s message is a grounded one.
Artificial intelligence is changing machine vision, but its greatest value comes when it works alongside the proven techniques and manufacturers with years of expertise.
