A literature review was conducted on image inpainting and image instance segmentation. The research gave the necessary theoretical foundation for the project. To have a systematic approach to develop and evaluate the developed prototype, the Design Science Research Methodology (DSRM) and evolutionary prototyping were used. Therefore, at the beginning of the project, a thesis statement, research questions, and a problem- and requirements analysis were conducted. Furthermore, based on those specifications, a conceptual design was sketched. Finally, the prototype implementation was started.

The prototype was implemented as a web application consisting of a user interface (Frontend) and a Backend, which implements the image segmentation and the inpainting. The components are connected via a Representational State Transfer (REST) Application Programming Interface (API). The Backend is based on FastAPI, a Python framework for web applications. Using Python provides access to many deep-learning libraries. The prototype incorporates the deep-learning libraries Large Mask Inpainting (LaMa) for the inpainting and Segment Anything Model (SAM) for the instance segmentation.

After the prototype was developed and functional, it had to be evaluated to see whether it fulfilled the defined research questions. A qualitative image comparison, as well as a user evaluation, were chosen. The findings of those evaluations were then compiled and summarized. The main finding of the bachelor thesis is that AI can be used to effectively remove furniture from furnished room images and replace them with visually coherent textures, demonstrating the potential of AI to revolutionize the field of interior design.