3D-FUTURE: 3D FUrniture shape with TextURE

Huan Fu1   Rongfei Jia1   Lin Gao2   Mingming Gong3   Binqiang Zhao1   Steve Maybank4   Dacheng Tao5

1Alibaba-inc    2ICT.CAS    3University of Melbourne    4Birkbeck College, University of London    5The University of Sydney

News

Introduction

The vision community has put tremendous efforts into 3D object modeling over the past decade, and has achieved many impressive breakthroughs. However, there is still a large gap between the current state of the art and industrial needs for 3D vision. One of the major reasons is that existing 3D shape benchmarks provide only 3D shapes with dreamlike or no textures. Another imperfection is that there is no large-scale dataset in which 3D models exactly match the objects in images. These are insufficient for comprehensive and subtle research in areas such as texture recovery and transfer, which is required to support industrial production. We thus release 3D-FUTURE (3D FUrniture shape with TextURE) which contains 20,240 photo-realistic synthetic images captured in 5,000 diverse scenes, and 9,992 involved unique industrial 3D CAD shapes of furniture with high-resolution informative textures developed by professional designers. More details can be found in our report.

Highlights

Alibaba will continously enlarge the benchmark by providing more annotations and adding new features (e.g., sharing the professional designs) to serve for 3D acadamic studies.

High-quality Shapes, Informative Textures, Rich Attributes

The 3D shapes offered by public 3D benchmarks may show two imperfections. Firstly, most of these 3D CAD models (for furniture) are both with fewer details and low informative textures since they are collected online. Secondly, there are no diverse professional attributes for their furniture shapes. In contrast, 3D-FUTURE provides high-quality 3D furniture with rich details in various styles, including European furniture that often contains intricate carvings. Furthermore, each 3D shape in 3D-FUTURE is assigned to an informative texture and different attribute labels. We believe these features can potentially facilitate innovative research on high-quality 3D shape understanding and generation.

Realistic Renderings, Real 2D-3D Alignment

There are no well-organized benchmarks that provide realistic synthetic indoor images. 3D-FUTURE fill the blank by rendering 20,240 photo-realistic synthetic images across 5,000 scenes via one of the most advanced industrial 3D renders (V-Ray). These indoor scenes are reviewed by professional designers. Besides, existing benchmarks only provide pseudo 2D-3D alignment annotations. Namely, they manually choose a roughly matched 3D CAD model from public 3D shape benchmarks according to the object contained in the image. Annotators thus may largely ignore some local shape details. As a result, these benchmarks offer less matched 3D shape and 2D image pairs. This is not sufficient to support data-driven studies such as high-quality 3D reconstruction and high-accuracy 3D shape retrieval. Luckily, the 9,992 3D shapes in 3D-FUTURE exactly match objects contained in the rendered images.

Download

The 3D-FUTURE benchmark is provided by Alibaba Topping Homestyler, and organized by Alibaba Tao Technology Department under the 3D-FUTURE Terms of Use. If you would like to download the 3D-FUTURE data, please apply for the dateset at TC Lab: 3D-FUTURE Dataset. If you have any questions, please send email to us at tianchi_bigdata@member.alibaba.com

.

Samples

  • All
  • model
  • Scene

Sofa Texture

Texture

Sofa Mesh

Mesh

Bed Texture

Texture

Bed Mesh

Mesh

Living Room

Render Image

Living Room Labels

Label Map

Bedroom

Render Image

Bedroom Labels

Label Map

Citation

If you use our dataset or code, please cite our report: 

    @article{fu20213d,
      title={3d-future: 3d furniture shape with texture},
      author={Fu, Huan and Jia, Rongfei and Gao, Lin and Gong, Mingming and Zhao, Binqiang and Maybank, Steve and Tao, Dacheng},
      journal={International Journal of Computer Vision},
      pages={1--25},
      year={2021},
      publisher={Springer}
    }