Roominoes: Generating Novel 3D Floor Plans From Existing 3D Rooms
Kai Wang, Xianghao Xu, Leon Lei, Selena Ling, Natalie Lindsay, Angel, X. Chang, Manolis Savva, Daniel Ritchie

TL;DR
This paper introduces a novel task of generating new 3D floor plans by recombining existing 3D rooms, aiming to enhance dataset diversity and facilitate applications in scene understanding and navigation.
Contribution
It defines the task, explores different solution pipelines, and proposes evaluation metrics for generating diverse and compatible 3D indoor scenes from existing data.
Findings
Two pipeline approaches demonstrated different trade-offs in subtask performance.
Metrics effectively evaluate layout, retrieval, and deformation quality.
Recombination strategies can generate diverse, realistic 3D indoor scenes.
Abstract
Realistic 3D indoor scene datasets have enabled significant recent progress in computer vision, scene understanding, autonomous navigation, and 3D reconstruction. But the scale, diversity, and customizability of existing datasets is limited, and it is time-consuming and expensive to scan and annotate more. Fortunately, combinatorics is on our side: there are enough individual rooms in existing 3D scene datasets, if there was but a way to recombine them into new layouts. In this paper, we propose the task of generating novel 3D floor plans from existing 3D rooms. We identify three sub-tasks of this problem: generation of 2D layout, retrieval of compatible 3D rooms, and deformation of 3D rooms to fit the layout. We then discuss different strategies for solving the problem, and design two representative pipelines: one uses available 2D floor plans to guide selection and deformation of 3D…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
