Artificial intelligence makes it possible for machines to recognize objects. This requires large amounts of high-quality image data with which the algorithms are manually trained. "Training AI systems to recognize objects is still time-consuming, inflexible, expensive, highly dependent on the environment and requires a lot of computing power," explains Kimoknow co-founder Lukas Kriete. The KIT start-up therefore uses image data that is already generated for all objects in computer-aided development processes (CAD) and production data management (PDM). Among other things, they provide information about the material, geometry and position of the respective object. The CAD and PDM data is extracted and used for the automated training of the AI.

The object recognition system trained in this way can be used in a variety of ways, including in augmented reality (AR) glasses. They detect relevant objects in the user's field of vision in real time and also have the necessary contextual information about the object in question.

Use case for AR glasses
Kimoknow has developed an assistance system as the first use case for AR glasses of this kind, which is designed to support skilled workers in the assembly of complex devices. The virtual assistant guides users through the entire assembly process, visualizes the assembly instructions step by step without an additional display and thus shows in which order which part is processed with which tools and assembly materials. It repeats individual steps if errors occur and documents the process. The fitter has both hands free and communicates with the system via eye contact, hand signals or voice commands. "The assembly assistant makes the process more efficient, more productive, faster and more cost-effective while improving quality," says Kriete.

The assistant is suitable for all industries in which highly complex products are manufactured in small quantities. The prototype is used for the final assembly of highly specialized measuring devices and is currently being tested in cooperation with Elabo at the Center for Artificial Intelligence Talents (CAIT) at the Institute for Information Management in Engineering (IMI) at KIT.

Kimoknow is a spin-off of IMI and was launched on May 13, 2020 as an entrepreneurial company (limited liability). In addition to Lukas Kriete, the founding team includes Roman Wiegand, Aaron Boll, Michael Grethler and Vesa Klumpp. as