S-INF: Towards Realistic Indoor Scene Synthesis via Scene Implicit Neural Field

TL;DR

S-INF introduces a novel scene implicit neural field approach that models multimodal relationships for more realistic indoor scene synthesis, outperforming existing methods on the 3D-FRONT benchmark.

Contribution

The paper presents a new method that disentangles scene layout and object relationships, integrating them via implicit neural fields to improve realism in indoor scene generation.

Findings

Achieves state-of-the-art results on 3D-FRONT dataset.

Effectively models multimodal relationships for realistic scene synthesis.

Ensures stylistic consistency across objects through differentiable rendering.

Abstract

Learning-based methods have become increasingly popular in 3D indoor scene synthesis (ISS), showing superior performance over traditional optimization-based approaches. These learning-based methods typically model distributions on simple yet explicit scene representations using generative models. However, due to the oversimplified explicit representations that overlook detailed information and the lack of guidance from multimodal relationships within the scene, most learning-based methods struggle to generate indoor scenes with realistic object arrangements and styles. In this paper, we introduce a new method, Scene Implicit Neural Field (S-INF), for indoor scene synthesis, aiming to learn meaningful representations of multimodal relationships, to enhance the realism of indoor scene synthesis. S-INF assumes that the scene layout is often related to the object-detailed information. It disentangles the multimodal relationships into scene layout relationships and detailed object relationships, fusing them later through implicit neural fields (INFs). By learning specialized scene layout relationships and projecting them into S-INF, we achieve a realistic generation of scene layout. Additionally, S-INF captures dense and detailed object relationships through differentiable rendering, ensuring stylistic consistency across objects. Through extensive experiments on the benchmark 3D-FRONT dataset, we demonstrate that our method consistently achieves state-of-the-art performance under different types of ISS.