FurniScene: A Large-scale 3D Room Dataset with Intricate Furnishing Scenes

Genghao Zhang,Yuxi Wang,Chuanchen Luo,Shibiao Xu,Zhaoxiang Zhang,Man Zhang,Junran Peng

2024-05-06

Abstract:Indoor scene generation has attracted significant attention recently as it is crucial for applications of gaming, virtual reality, and interior design. Current indoor scene generation methods can produce reasonable room layouts but often lack diversity and realism. This is primarily due to the limited coverage of existing datasets, including only large furniture without tiny furnishings in daily life. To address these challenges, we propose FurniScene, a large-scale 3D room dataset with intricate furnishing scenes from interior design professionals. Specifically, the FurniScene consists of 11,698 rooms and 39,691 unique furniture CAD models with 89 different types, covering things from large beds to small teacups on the coffee table. To better suit fine-grained indoor scene layout generation, we introduce a novel Two-Stage Diffusion Scene Model (TSDSM) and conduct an evaluation benchmark for various indoor scene generation based on FurniScene. Quantitative and qualitative evaluations demonstrate the capability of our method to generate highly realistic indoor scenes. Our dataset and code will be publicly available soon.

Computer Vision and Pattern Recognition,Artificial Intelligence

What problem does this paper attempt to address?

The paper primarily addresses two key issues in indoor scene generation:

1. **Lack of Diversity and Realism**: Existing indoor scene generation methods can produce reasonable room layouts, but they are limited in terms of diversity and realism. This is mainly because current datasets have limited coverage, often including only large furniture and lacking small decorative items found in everyday life.

2. **Dataset Limitations**: The datasets currently in use are limited in terms of variety and scene complexity, such as having a small number of furniture types, resulting in monotonous scenes that differ significantly from real-life scenarios.

To address these issues, the authors propose FurniScene—a large-scale 3D room dataset with the following features:

- Contains meticulously designed indoor scenes created by professionals, ensuring high aesthetic quality.
- Large data volume, including 111,698 rooms and 39,691 unique furniture CAD models, covering 89 different types of objects.
- Diverse room layouts, with each room containing an average of 14.4 objects, and up to 119 objects.
- Rich scene details, with each room being very complete and including many decorative items.

Additionally, to better handle the task of generating detailed indoor scene layouts with numerous decorative items, the paper proposes a new model called the Two-Stage Diffusion Scene Model (TSDSM) and conducts benchmark evaluations based on FurniScene. The results show that this method demonstrates superior performance on multiple quantitative metrics.

In summary, this research aims to enhance the realism and diversity of indoor scene generation by constructing a high-quality, large-scale dataset and proposing an effective generation model, thereby advancing research progress in the related field.