Machine vision is one of the most demanding fields in the world of information technology

Over the last few decades, MVTec Software GmbH has developed from a spin-off of the Technical University of Munich to one of the world’s leading software companies for machine vision. Ever since its establishment, MVTec has been setting standards with regard to the quality of its software products.

Today, an around 80-person software development team is dedicated to offering customers the most powerful and sophisticated algorithms. In an interview, Johannes Dieterich, Head of Engineering, and Andreas Lösel, a software engineer in the deep learning team, reveal what makes software development at MVTec special, which topics the developers focus on and what the prerequisites are for working for MVTec.

What makes MVTec so special for software developers?

Andreas: For me, it’s the combination of several aspects. On the one hand – and this is probably the most important aspect – I work on and with state-of-the-art technologies every single day. At the same time, our projects aren’t focused on a specific area. Instead, our software is used in completely different projects and in practically all sectors. On the other hand, the company encourages my personal and professional development. Another aspect, which is also important to me, is that we have an extremely positive working atmosphere where everyone respects one another.

Johannes: What makes us different from other companies is that our work involves highly demanding, technology-related tasks. Our core discipline naturally lies in machine vision, but our fields of work go far beyond that. For example, we’re creating a proprietary development environment that is optimized for machine vision together with appropriate programming language. As a computer scientist, I find these truly fascinating professional topics. Another factor is that our team is characterized by great cohesion and a very special spirit. As a medium-sized enterprise, we thrive on cooperation and short communication channels. These are essential aspects for our continued success.

Where exactly are MVTec’s software products used?

Johannes: Where aren’t they? There are few economic areas where machine vision can’t help improve performance or efficiency. In the industrial environment, we’ve always been represented in the mechanical engineering, automotive production, and electronics industries. Machine vision is also used in the logistics industry and even in the agricultural sector. At present, we’re particularly focusing on the semi-conductor industry and battery production. All over the world, major production capacities are currently being developed for these future-oriented sectors – and machine vision is a key aspect for production. On the other hand, we’re being more restrained with regard to the fields of armament and surveillance, as these are incompatible with our mission statement.

Andreas: As a developer of deep learning methods, the most important thing for me is which tasks can be resolved through my work. This is always the case where complex patterns make detection particularly challenging due to the difficulty of defining clear rules. Examples include anomaly detection, for example for quality inspections or OCR, i.e., capturing text in images automatically.

Why do companies opt for MVTec’s software? Or to put the question another way, how and with what does MVTec impress customers?

Andreas: I dare to say that we’re able to offer the perfect package of user friendliness and technological excellence. We’re committed to combining classic machine vision with state-of-the-art technologies in the field of machine learning.

Johannes: MVTec’s founders met at the Technical University of Munich and their thirst for research and aspirations with regard to technology still shapes our actions today. For example, we have our own research department, where we work with the latest scientific findings and analyze whether and how we can use these to deliver added value to our customers. This gives us a competitive advantage with regard to technology. One specific example is the deep-learning-based technology Global Context Anomaly Detection developed by MVTec. This is a world first in this form. At the same time, the technology also acts as an example of how important deep learning has become in the world of machine vision. We’re at the very forefront of this development.

What specialist background do MVTec developers have?

Andreas: I personally studied electrical engineering and information technology. When completing my master’s degree, I could choose my subjects quite freely and focused on machine learning. However, there’s no rigid path that you have to take to work for MVTec. For example, everyone in my team comes from a different background.

Johannes: Machine vision is a very demanding field of information technology, possibly more so than other areas of use of software. While you definitely need good programming skills to work at MVTec, those alone are not enough. For example, our MVTec HALCON software is one of the most powerful machine vision software products on the market. It contains over 2,100 operators, is accordingly complex, and includes several million lines of code. Working with this not only demands the highest level of knowledge in algorithmics, but also in software engineering. Deep learning is yet another world of its own. That means that university graduates with deep learning expertise won’t automatically end up working in the field of machine vision. Here at MVTec, we’ll therefore have to contend with more competitors on the labor market than we used to. What gives us an edge here is that we’ve already been working intensely with deep learning for a long time. Graduates can therefore benefit from the chance to learn from recognized pioneers and experts in the field.

If you listen to public discussions, you get the impression that Germany is lagging behind when it comes to AI. But this doesn’t seem to apply at MVTec?

Andreas: Unfortunately, technological progress is often only reported (and then widely so) in the case of sensational breakthroughs. We continually make innovative advances, but these aren’t particularly spectacular. Despite this, they’re still groundbreaking for our sector.

Johannes: Exactly. Although machine vision is a core component with regard to automation, it STILL remains somewhat of a niche topic. There’s also the added factor that economic issues generally receive less attention than those that affect consumers in general. We don’t need to hide from our international competitors with regard to technology, including in terms of deep learning – quite the contrary.