SceneNAT: Masked Generative Modeling for Language-Guided Indoor Scene Synthesis

TL;DR

SceneNAT introduces a masked non-autoregressive Transformer that efficiently generates complete 3D indoor scenes from natural language, outperforming previous methods in accuracy and computational efficiency.

Contribution

The paper proposes SceneNAT, a novel masked modeling approach with a triplet predictor for improved scene synthesis from language instructions.

Findings

Outperforms state-of-the-art methods in semantic and spatial accuracy

Operates with significantly lower computational cost

Effectively captures intra-object and inter-object relationships

Abstract

We present SceneNAT, a single-stage masked non-autoregressive Transformer that synthesizes complete 3D indoor scenes from natural language instructions through only a few parallel decoding passes, offering improved performance and efficiency compared to prior state-of-the-art approaches. SceneNAT is trained via masked modeling over fully discretized representations of both semantic and spatial attributes. By applying a masking strategy at both the attribute level and the instance level, the model can better capture intra-object and inter-object structure. To boost relational reasoning, SceneNAT employs a dedicated triplet predictor for modeling the scene's layout and object relationships by mapping a set of learnable relation queries to a sparse set of symbolic triplets (subject, predicate, object). Extensive experiments on the 3D-FRONT dataset demonstrate that SceneNAT achieves superior performance compared to state-of-the-art autoregressive and diffusion baselines in both semantic compliance and spatial arrangement accuracy, while operating with substantially lower computational cost.