Hanan Tanasra is an Architect and an MSc student at the Faculty of Architecture and Town Planning at the Technion.
Research Topic: Automation in Design
A Comparative Study of CGAN Models for Interior Room Generation
The furnishing stage in interior design involves a series of iterative adjustments to fulfill design objectives, incorporate professional input, and optimize design performance. However, with the advent of Machine Learning (ML), there is a potential to automate and improve the residential design process while maintaining creativity and quality. In our research, we propose a furnishing method that utilizes ML to streamline the design process and enhance the overall interior design experience. We trained and evaluated three CGAN models – Pix2Pix, BicycleGAN, and SPADE – on bathrooms. The results showed that BicycleGAN displayed better performance according to our ML evaluation criteria. Visual analysis confirmed that models generated floor plans adhering to architectural standards. The proposed method can suggest new design options generated in a short time while eliminating the manual work of furnishing the desired design. This allows designers and non-designers to create good interior designs despite their lack of architectural understanding.