Arranging Furniture Using Machine Learning in Virtual Environment

With the advancement of virtual environment technologies, solutions for arranging and visualizing furniture layouts in 3D environments based on user preferences are becoming increasingly popular. This thesis investigates the possibility of using a language model to generate IKEA furniture arrangement solutions on a 3D simulation platform. The study begins by integrating human feedback into language model training to enhance its capabilities in furniture layout design. It further explores the interaction between the language model and a 3D simulation platform, using generated outputs to create virtual furniture arrangements based on user prompts. The study also examines the feasibility of importing IKEA furniture models into virtual environments to produce realistic layout solutions. The results reveal a noticeable gap in quality between the scenes generated by the 3D simulation platform using LLMs and those designed by IKEA. However, the study underscores the promising potential of integrating natural language processing with 3D scene generation for automated, user-centered furniture arrangements.