Showroom

Showroom with AI demonstrators

Eight AI demonstrators were developed as part of the "SDaC" research project (funded by the BMWK, project duration 04/2020 - 09/2023).

These demonstrators are intended to demonstrate the potential of AI in construction planning and realization and to inspire software companies to add intelligent components to their products.

Elements of human-machine collaboration were taken into account in all demonstrators. In this way, the accuracy of the forecast can be improved and confidence in the solutions increased at the same time. The demonstrators show that despite the hyped term "artificial intelligence", the construction industry will not be able to do without you, the experts.

The demonstrators include common construction industry data formats (e.g. JPEG, PDF, Excel) that require intelligent processing. The demonstrators were trained with a data set provided by the consortium and can therefore only be transferred to internal company data to a limited extent - the data formats or the trained classes must therefore be taken into account.

Enclosed you will find files that you can use to test the applications. These can be downloaded as a zip archive.

Object detection in construction plans

Construction project data is often only available in the form of paper plans or outdated CAD models. Currently, the structures required to use this data as a basis for BIM models are lacking. In this demonstrator, we use so-called "Convolutional Neural Networks" to automatically process this input data using the example of electrical symbols and to recognizeand label objects. The work focuses in particular on the meaningful design of the interaction between domain experts and AI models

Developer: Karlsruhe Service Research Institute (KSRI) & digitales bauen GmbH (as part of Dress & Sommer)

Image classification

Many of the images taken on construction sites are stored on project drives without being named. This causes a great deal of manual effort when searching for specific objects. In this demonstrator, images are therefore automatically assigned to one of 35 object classes. A semi-supervised learning approach was used to reduce the effort required for "labeling". This demonstrator can also be obtained as an API on request in order to integrate it into existing software products.

Developer: Karlsruhe Service Research Institute (KSRI) & digitales bauen GmbH (as part of Dress & Sommer)

Layout generator

The layout generator is used to create floor plans of a single or multi-storey building. This requires the building contour and a room list in which various boundary conditions for the subsequent floor plans can be defined. One or more optimized floor plan layouts are calculated from the information and displayed visually. The floor plans created can be exported in the openBIM data format IFC and used in compatible software tools for further planning steps. The demonstrator is a Windows desktop application.

Developer: Institute for Automation and Applied Computer Science (KIT)

Interested in more?

On the platform developed during the project, you will find further use cases for the classification of defects in the non-uniform naming of descriptions of defects, the supplementation of a bill of quantities by means of a database on costs and durations (schedule preparation), the analysis of potential supplements based on a bill of quantities, the digitization of delivery bills and the recognition of construction progress as well as hazard detection in images.

Platforms in the context of AI developments support joint learning through the sharing of data and user feedback.

Developer: Fraunhofer ISST

AI tools for the construction industry are not unique to us.
Get an overview of software solutions for the construction industry that include AI elements.

The following economic applications
have emerged from SDaC, among others...

Valoon

Valoon combines the advantages of construction project management software with the simplicity of messaging services.

Delivery note detection