Timeframe and Context

The project was carried out from October 2023 to February 2024 as part of a master’s degree project in cooperation with the Allgäu Research Centre and a company based in Kempten.

The goal was to automatically segment components on panel layout plans and project the construction plan onto the work area to increase employee efficiency.

As a developer, I was responsible for preprocessing the plans, calculating the actual size, detecting duplicates, and recognizing standard tiles.

Implementation and Tech Stack

The implementation included preprocessing the plans, segmenting and classifying all objects, and postprocessing the results. The goal was to accurately recognize the components and correctly display the target position of the current component in the work area.

Python and SAM (Segment Anything Model) were used for segmentation.

Challenges and Results

The diversity of the components, the varying scope and complexity of the plans, and a limited database posed key challenges.

The solution was able to reliably recognize over 98% of the components. In addition, the target position of the components is displayed, creating a solid foundation for subsequent automation and efficiency improvements.