An Empty Room is All We Want: Automatic Defurnishing of Indoor Panoramas

TL;DR

This paper introduces a pipeline using Stable Diffusion for defurnishing indoor panoramas, achieving high-quality, geometrically plausible inpainting without room layout estimation.

Contribution

It presents a novel defurnishing method leveraging domain-specific fine-tuning and advanced blending to improve inpainting quality over existing techniques.

Findings

Qualitative and quantitative improvements over prior methods

High-fidelity inpainting without room layout estimation

Effective use of Stable Diffusion for indoor panorama editing

Abstract

We propose a pipeline that leverages Stable Diffusion to improve inpainting results in the context of defurnishing -- the removal of furniture items from indoor panorama images. Specifically, we illustrate how increased context, domain-specific model fine-tuning, and improved image blending can produce high-fidelity inpaints that are geometrically plausible without needing to rely on room layout estimation. We demonstrate qualitative and quantitative improvements over other furniture removal techniques.