As part of TÜV inspections, weld seams and cracks on shipboard container cranes must be inspected. To do this, an on-site inspector captures a large number of images, all of which then need to be reviewed and assessed afterward. The previous inspection process involved a very high level of manual effort. Every recorded image had to be reviewed individually, even though only a portion of them actually showed critical cracks, relevant welds, or concrete areas of concern. The goal was to automatically filter the relevant content from the collected images, reliably identify suspected defects, and significantly reduce the number of images requiring manual follow-up review. At the same time, the inspector’s existing workflow needed to remain unchanged. With Maddox AI, the captured images are automatically pre-sorted. Relevant cracks, welds, and suspected defects are filtered out in a targeted way, so inspectors only need to review the images that are actually relevant for assessment. This preserves the existing process while significantly reducing overall inspection effort.