HouseLayout3D: A Benchmark and Training-Free Baseline for 3D Layout Estimation in the Wild

TL;DR

HouseLayout3D introduces a real-world benchmark for full building scale 3D layout estimation, addressing limitations of existing models trained on synthetic single-room data, and presents MultiFloor3D, a training-free baseline that outperforms current methods.

Contribution

The paper provides a new benchmark for multi-floor 3D layout estimation and introduces MultiFloor3D, a training-free baseline that surpasses existing models on this benchmark and others.

Findings

MultiFloor3D outperforms existing models on the new benchmark.

Existing models struggle with multi-floor, architecturally complex environments.

The benchmark enables progress toward full building scale 3D layout estimation.

Abstract

Current 3D layout estimation models are primarily trained on synthetic datasets containing simple single room or single floor environments. As a consequence, they cannot natively handle large multi floor buildings and require scenes to be split into individual floors before processing, which removes global spatial context that is essential for reasoning about structures such as staircases that connect multiple levels. In this work, we introduce HouseLayout3D, a real world benchmark designed to support progress toward full building scale layout estimation, including multiple floors and architecturally intricate spaces. We also present MultiFloor3D, a simple training free baseline that leverages recent scene understanding methods and already outperforms existing 3D layout estimation models on both our benchmark and prior datasets, highlighting the need for further research in this direction. Data and code are available at: https://houselayout3d.github.io.