experimenta is a hands-on science centre in Heilbronn, officially recognized by the State of Baden-Württemberg as an extracurricular learning location and research center.

Jointly with IDS who provided cameras, we created AI-enabled vision for a station where visitors can experience being newspaper editors and create newspapers with custom layout and content. The station uses wooden blocks for layout and bricks with data matrix codes for content. These are placed into a special grid and the resulting newspaper design appears on the screen.

Our vision solution detects the newspaper configuration with a smart camera located under the grid. This is difficult to achieve as the camera also detects people, clothes, hands, bags, etc and in a variety of lighting conditions (e.g. daylight, artificial light).

We used deep learning – the most advanced technology available – to achieve precise object recognition (our product Aurora). Urobots were also ahead of the times, pioneering the launch of AI-enabled vision on an early generation camera, which was not designed to work with AI.

Highlights

• Brought deep learning software into low-energy consumption CPU (in contrast to high-energy GPUs typically used for this purpose).
• Unlike most industrial applications, our solution can work in an ‘unpredictable’ environment, i.e with moving people and changing light conditions.
• In addition to deep learning, created algorithms to increase the reliability of object recognition to nearly 100%.