A key aspect of IIMEO is the AI-assisted detection of obstacles on the tracks. To achieve this, the AI model must be trained to recognise tracks in image data.

A flight campaign had originally been planned for the autumn, but this could not be carried out due to poor weather conditions and time constraints. Following consultation with the team at Miniatur Wunderland in Hamburg, Germany, we were offered the opportunity to relocate our ‘flight campaign’ to the model landscapes at Miniatur Wunderland and to capture images there across various landscapes and featuring different obstacles as examples of potential disruptions.

Last December, we visited Miniatur Wunderland on three evenings to collect our reference data for the model training. Although the environment at Miniatur Wunderland did not quite correspond to reality, it nevertheless provided us with the opportunity to capture a variety of valuable training data regardless of the weather conditions – an efficient solution despite the challenges.

A big „Thank you““ goes to the Miniatur Wunderland team, who provided us with excellent support on these occasions.