Kinetic Vision
Trailblazers in Product and Process Development

Despite a slow start, augmented reality (AR) and virtual reality (VR) have become a disruptive force in business. Though AR and VR have been around for more than three decades, only recently have these technologies moved beyond consumer games and entered the commercial space. However, the road leading AR and VR to success in the business world is still slowed by major challenges, including a lack of sufficient resources and knowledge, and difficulty in defining use-cases with a measurable return on investment (ROI). This is what Ohio-based Kinetic Vision, a full-service research and development firm, strives to address.

Kinetic Vision has helped leading companies integrate the latest technologies in their product development, manufacturing, and quality processes for over thirty years, and is proud to claim over 50 of the Fortune 500 as customers. Its key differentiator is the ability to bring a diverse set of skills to its projects, including industrial and mechanical design, embedded hardware and electrical engineering, modeling and simulation, full prototype capabilities, software and mobile development, machine learning, and AR/VR/ XR expertise.

Recently Kinetic Vision established a new division, Deep Vision Data, to fulfill the growing demand for deep learning solutions and the data needed to train them. That led to the development of aiVisionTM, a platform for creating virtual environments to train, deploy and test Artificial Intelligence (AI) systems.

In an interview with CIO Applications, CEO Rick Schweet shares insights on the current AR/VR landscape, the company’s unique value proposition, along with the benefits of the aiVision platform.

How has the AR/VR market evolved over the years, and how has Kinetic Vision maintained its leadership?

Although functional VR systems were developed decades ago, hardware costs and slow graphics processing limited its usefulness to R&D and specialized high-end applications. AR is a more recent technology that’s recently become available to the masses due to the ever-increasing power of mobile devices. The first wave of applications focused on consumer use: games, improved shopping experiences, augmented street navigation, etc. We are bringing these same technologies to product R&D, visual marketing, and training.

Kindly walk us through the multidisciplinary skills of Kinetic Vision and why you feel they are important.

A decade ago, the vast majority of products were either electronic or mechanical. An electric toothbrush was mechanical; the head vibrated and that was about it. Now electric toothbrushes connect to apps that tell you how well you’re brushing your teeth, when you need to replace the head, and much more. Soon everyday devices will use machine learning to understand your habits and needs almost better than you do.

Developing the smart products of tomorrow requires skill in many disparate areas. Using the toothbrush example again, you need industrial and mechanical design, mechanical engineering, electrical design and engineering, embedded development and hardware design, and even machine learning expertise. To prototype your product, you need injection molding machines and on-demand printed circuit board manufacture and assembly. Most companies specialize in a few of these areas, but I’ve yet to find another company that provides all these capabilities under one roof—we do. And for complex systems, we’re doing more of the development in a virtual environment.

Could you shed light on the company’s deep learning platform, aiVision?

Many recent successes in AI, such as self-driving vehicles, voice processing, and healthcare, use a technique called deep learning. Deep learning utilizes computer models that roughly emulate the neural networks of the human brain. They are trained by providing the model with a very large number of examples of the items of interest. That’s easy when you’re looking for cats in photos because there are millions of those images online, but what if you’re developing an autonomous vehicle? You need billions of driver-miles of camera and sensor data, and creating that data with test fleets could take hundreds or even thousands of years.

This problem can only be solved with synthetic data, that is, data that the AI model interprets as real but is created artificially. That is the approach that Waymo and many others are taking.

We were early to recognize the effectiveness of synthetic training data for deep learning, and two years ago created a separate division called Deep Vision Data to address this market. Although the types of data we create vary widely, all deep learning training data have a common requirement: the need to control the data feature variability and distribution. To address this common need we created a development platform called aiVisionTM. For example, teaching an autonomous vehicle to drive down the center of a well-marked highway lane is relatively easy using physical data because many drivers spend a high percentage of their time doing this and there is ample data; but having it understand that a snowman on the side of the road isn’t going to walk out in front of it is much, much harder. That’s because your test fleet of vehicles collecting physical data might never encounter a snowman, and certainly not thousands of examples. Our aiVision platform makes this easy: you just set the “dials” to specify how many snowmen you want to appear in the virtual scenes.

Could you provide a customer success story of Kinetic Vision’s offering?

Recently, a well-known company worked with us to develop a complex robotic system. This system functions in a specialized environment in collaboration with other equipment and technicians, and the interactions are extremely important. Traditionally systems like this were developed with physical prototypes, but when you’ve got rooms, robots, and equipment that quickly becomes slow and prohibitively expensive. Our approach was to do the development entirely virtually using something called Extended Reality or XR. XR is similar to VR in that is includes an immersive 3D environment, but it adds virtual mechanical systems, computer software, sensors, cameras, and other functionality. By utilizing XR we were able to quickly optimize the system’s mechanical design, control software, and room configuration, while also collecting user feedback interactively. This approach probably saved our customer years of development time and millions of dollars, and because we could iterate quickly, ultimately resulted in a better product.

Another example is a camera-based system for retail we developed as a demonstration of our technology. Amazon Go is the best-known example of a cashier-less shopping experience, but it was reportedly developed using a physical store prototype, with real products and human shoppers. People familiar with the project estimated that Amazon invested hundreds of millions of dollars and seven years in that effort; I believe our aiVision platform could create the same system at 5 percent of the cost in 15 percent of the time.

What is the secret behind Kinetic Vision’s success?

Our success is directly linked to three main things: our skillset diversity, ability to move fast like a startup, and stable workforce. As I previously mentioned, the smart products and AI systems of today require the skills and collaboration of many different disciplines. We provide “one-stop shopping” for our customers and that saves them considerable time and money. Startups get a lot of attention, but the reality is that 75 percent of them fail within five years. We move fast like a startup but have over 30 years of growth and success, so we can form long-term relationships with our customers. Finally, because of the nature of our business, we hire people that are in very high demand, such as data scientists, mobile app developers, AR/VR developers, etc., and retaining that talent is an issue for most companies. We’re fortunate that our annual staff turnover is less than 2 percent, which is extremely low for our industry. That creates consistency in our client relationships, and our customers appreciate that.

What does the future hold for Kinetic Vision?

AI already influences many things we do day-to-day. We ask Alexa to help us with activities and information, we use ride-sharing apps instead of taxis, AI systems recommend movies, products and music we might like, our car navigation intelligently routes us to our destinations, and new use-cases appear nearly daily. Very soon AI will help professionals like lawyers and radiologists do their job better, and will eventually replace many of them. The next frontier is product development, quality control, and manufacturing. We’re already experts in those fields so we understand what the tough problems are. Our future is helping companies utilize the power of AI to innovate faster at a lower cost, and we’ve already had great success in that area.