Of error devils and AI: 42 intern develops new solution

May 2023: Error message at camera manufacturer IDS. Several cameras in a large order of around 600 are defective. The error pattern: diffuse. How many are affected is still unclear. Time is of the essence: as many cameras as possible must be delivered to the customer within two weeks, and they must be free of defects. Tim - a student at 42 - develops a performance check. And the delivery leaves the factory in Obersulm on time.

IDS builds industrial cameras. More than 350 employees, 11,000 customers in 70 countries, around 1.8 million IDS cameras in use worldwide - the term “hidden champion” was invented for companies like IDS. The cameras and image processing systems are used in equipment, plant and mechanical engineering. They are also used in agriculture and medical technology. Managing Director Jan Hartmann: “Even if it sounds banal, it is quality and innovative strength with which we score points worldwide. Sending out defective cameras wouldn't work for long.”

Trouble-shooting department

Tim works as a trainee in the very department that is consulted in the event of quality problems. The 42-year-old: “We are actually the trouble shooters at IDS. If there's a hitch somewhere in production and the software used in production is causing problems, our phone rings.” It's important to know that IDS manufactures highly specialized devices with more than 3,000 different models.

Tim's trouble-shooter quality is put to the test early on at IDS. Tim: “From day one, I was supposed to develop new software that could identify complex sources of error in production and help to improve production. Then came the aforementioned problem with the large order and my boss asked if I could program a performance check within a week.” Tim can. He finished the software on a Friday, one day before the deadline. And it works. Defective cameras are reliably detected and sorted out, IDS delivers on time. “I thought, how cool is it that my coding ideas can be used so directly,” says Tim.

The fact that it worked out so well also has a lot to do with the 42 spirit. Jan Hartmann points out that the coders at 42 are characterized by a very special way of working: “Compared to traditional university students, they approach problems differently. There is an enormous amount of independence and they combine the 42 coders with a pronounced team spirit - impressive.”

AI showcase

These skills also qualify Tim to support IDS in the field of AI. The company is one of the pioneers and began exploring the potential of AI for camera systems five years ago. Hartmann: “If camera systems are brought together with the right AI image processing, great things are possible.” As an example, he refers to an agricultural application: “Everyone wants to use less pesticides. If you install an AI-supported camera system on a tractor or sprayer, the pesticides can be applied precisely to the weeds, while large areas of arable land remain untreated.”

A promise, also for countless industrial applications. However, according to Hartmann, there are major reservations among SMEs in Baden-Württemberg in particular. AI is difficult to bring to the people. There is too much concern that the much-vaunted AI black box could cause damage. The IDS Managing Director sees himself as a bit of an educator here. His approach is to show AI applications in concrete terms, explain a lot and allay fears. With this in mind, an AI-supported production system can be inspected in Obersulm. Hartmann: “We are currently developing our own AI application and will soon be presenting it as a kind of showcase. The open exchange about limits and possibilities is extremely important.”

Full-time

Tim's job is to train the AI software for the production of a new camera model. The goal: to control a production step with the help of AI and prevent the high-quality individual parts from being assembled incorrectly. Tim: “Together with my colleagues, I came up with all kinds of scenarios of what could go wrong in production.” To do this, he took hundreds of images, classified them as correct and incorrect and fed them to the AI. The fact that Tim had no previous experience with AI didn't bother anyone at IDS. Incidentally, Tim is staying at IDS a little longer: from September, he will no longer be an IDS intern, but a full-time employee!