← Back to Computer Vision
cs.CV

Can AI design whole homes from empty floorplans?

Wenbo Li, Xiaoliang Ju, Zipeng Qin, Rongyao Fang, Hongsheng Li

June 4, 2026

Generating complete, realistic home interiors is hard—existing methods tackle floorplans or single rooms in isolation, missing global coherence. This work chains together a language model (trained on 300K real floorplans) to generate layouts, image diffusion models to place furniture from multiple viewpoints, and a vision-language model to iteratively fix placement errors. The result: fully furnished, simulation-ready homes with diverse layouts and high design quality. The team releases their floorplan dataset and 5,000 complete scenes.
Published as HomeWorld: A Unified Floorplan-to-Furnished Framework for Generating Controllable, Densely Interactive Whole-Home Scenes arXiv:2606.06390
Read the original paper →