November 17, 2020
At Wikitude, we are proud to say that we have a talent to find (& keep) talent. To show you how, we are introducing our new interview series, In Focus. Each article will tell a story about a team member that makes Wikitude what the company is today. Let us introduce Markus Eder, Head of Computer Vision team, who has celebrated nine years at Wikitude.
What made you take the leap and start your AR/Computer Vision journey?
When I studied computer science for my Bachelor’s degree, many things started shifting in the industry. The first smartphones were just introduced, with their new capabilities (especially these of iPhones), allowing a whole new category of applications.
A concept that immediately caught my attention was an app called London tube. It overlaid the location of the next tube station in London over the running camera feed. I liked the concept and its potential and started to read about the technology and theory behind it. This search first brought me to augmented reality and further on to the whole area of Computer Vision (CV). So much so that I decided to focus my Master’s on it.
After doing a semester abroad in Australia, I looked for opportunities to get a paid Master’s thesis. Such an opportunity arose in Salzburg at a local research center. There I developed a concept for AR assisted pedestrian navigation on a mobile device. To improve the user experience, it combined Geo AR with Computer Vision algorithms. As the thesis was completed, I realized that I want to stay in this field.
At the time, the job openings in Computer Vision were scarce, but luckily I landed one directly in Salzburg – at Wikitude.
What were your initial expectations?
When I joined the company, I thought I could continue in the same area as my thesis. But soon, I realized there is a clear gap between scientific research and product development. One thing is to show a prototype proving that a particular technology works. Developing that prototype into a ready-for-market product that works under all circumstances proved to be a whole other topic.
From a research perspective, those days were really exciting. Many Computer Vision concepts and groundworks that are applied in current AR frameworks were conceived around that time. As research topics like SLAM, Structure-from-Motion, or 3D Reconstruction were evolving, it became apparent that these ideas will enable a new generation of AR capabilities (even on a mobile phone). Wikitude, Imagination and Metaio (now part of Apple) were the first movers to integrate some of these ideas into their products.
What challenges or setbacks did you face along the way?
In 2012, a few AR apps in the market used the device’s sensors (GPS, Accelerometer, Compass) solely to overlay AR content on the phone’s camera stream. Back then, the focus was to market the existing solution rather than improving technology. So the first technical challenge was shifting from sensor-based to Computer Vision-based algorithms for AR visualization. But once we did, it opened the door for an entirely new category of AR use cases which have shaped the market as we know it today.
A second challenge was playing in a highly competitive environment. Up until now, the leading AR companies are far larger. Despite that, we’ve been competitively leading the space. I believe this drives all Wikituders to compete with those companies and offer better features and quality.
How did Wikitude transition from an app to an AR SDK provider?
When Wikitude was still a young startup, the atmosphere encouraged the exploration of new ideas. It facilitated the integration of new features into the application. Soon, it got clear that there is a massive potential in allowing other developers to integrate AR technology into their products. This shift in focus meant changes in the tech teams to optimize our resources and develop a full-merged AR SDK.
As you progressed, did any part of your journey change? How?
Initially, the focus lay mainly on the Geo AR-based Wikitude app, so I worked on the Android side. The more we realized CV and AR’s potential, the more my focus shifted in that direction. At this point, we decided to create a new team to focus on R&D in that area solely. For me, it meant that I could work on the CV-based research aspects of the SDK. At the same time, we conducted state-funded research projects with several Austrian universities, which are among the best in AR and computer vision. The research results and expanding our team with international hires have helped us a jump-start in the right direction.
Can you share any tips on how to build a successful CV team?
With an increased demand for Computer Vision-based features in the SDK, we continued to hire more people. It was hard to find people with a computer science background in general and with a particular skill set. Quickly we realized that we have to look for potential candidates internationally as it still was a very specialized field.
As a tight-knit collective, we can not afford to hire the wrong people. I guess I always go with my gut feeling, and time shows it has been working.
There is also an exceptional working atmosphere in Wikitude where each team member feels that they can bring something to the table, and it encourages people to do their best.
How were your expectations met along throughout your journey at Wikitude?
When I think back to where we started, the whole journey exceeded my expectations – especially considering the market changes in the last couple of years. After all these years, we are one of the leading AR technology providers in the market with a vast customer base. The profile of our customers changed drastically over the years.
In the early stages of our SDK, customers were most interested in creating a “wow”-effect for the customer by showcasing necessary information in AR. This approach has changed – now, the customers come with business cases in mind where AR has a clear benefit for the users. Just showing something in AR is not enough. Relying heavily on constant communication with our clients, we have significantly changed our offering and the feature set we provide to the developers.
Did you achieve what you wanted to?
After completing my studies, my goal was to deepen my knowledge and continue to work and research in that field. Since then, AR has progressed so far that it influences people’s lives and assists in daily use cases.
Until the technology hasn’t reached its full potential yet, I will not stop working in this field until it’s done.
From your perspective, how is the future of AR and Computer vision will look like?
I believe that in the future, we’ll move away from the web-based AR experiences where users have to hold a device in their hands as it restricts the use cases and interaction. As technology evolves, we’ll see a more natural and less intrusive way to interact with augmented reality content that is comfortable and intuitive for users.
Another aspect that will change is the level of immersion. At the moment, in most cases, users statically look at augmented experience. Along with the hardware that will handle heavy computing and advanced optics, augmented experiences will evolve and offer more advanced graphics and interaction. We can already see how devices have solved the problem of localization, next up will be detecting and recognizing the objects in user’s environment to create a context.
As a company, we are working closely on solving these challenges to support more use cases and make the engine smarter to recognize more complex objects.
What would be your advice for professionals who want to enter and succeed in the area of Computer Vision and Augmented reality?
Many ways could lead to this area. Having a solid background in advanced mathematics will help along the way to understand the fundamental concepts of computer vision and AR. It has become more accessible in the last couple of years, thanks to dedicated programs focused on computer vision and the progress of deep and machine learning.
Another helpful thing is defining which specific area you’re interested in – whether it’s visualization (an essential aspect of AR) or other areas, which comprise the fundamentals of computer vision and machine learning. There is much research happening in this area, but trying your hands on the actual technology could be very useful.
Get more insights about our team: Read the interview with Wikitude COO Nicola Radacher