SPG-Layout generates physically plausible indoor scenes in non-Manhattan environments

Researchers propose SPG-Layout, a text-driven framework designed to generate physically plausible 3D indoor scenes within complex non-Manhattan environments where existing methods often fail due to high geometric violations.

The approach utilizes statistical priors of object distributions to enhance environmental understanding and adopts a hierarchical layout strategy that prioritizes large objects to minimize layout violations. The authors constructed a new benchmark comprising 500 diverse non-Manhattan environments to evaluate performance in these complex settings.

Extensive experiments demonstrate that SPG-Layout consistently and significantly outperforms existing methods across both Manhattan and non-Manhattan environments, achieving a balanced optimization of semantic realism and physical plausibility.