3D-FUTURE: 3D Furniture shape with TextURE

TL;DR

3D-FUTURE tackles the shortage of richly textured, fine-grained 3D furniture data for household interiors by introducing a large-scale, richly annotated benchmark with 9,992 textured CAD models and 20,240 photorealistic scenes across 5,000 rooms. It combines an automated Furnishing Suit Composition framework (DFSM) with a deep visual embedding and rule-based reasoning to facilitate efficient room design, plus integration with the Topping Homestyler platform for practical usability. The authors provide extensive baselines across 2D-3D retrieval, reconstruction, pose estimation, segmentation, and texture synthesis to demonstrate dataset versatility and to identify current limitations in high-detail 3D understanding. The dataset aims to bridge academic research and industrial production, enabling advances in texture recovery, fine-grained geometry, and interior understanding. Overall, 3D-FUTURE offers a versatile resource and concrete baselines to spur progress in high-quality 3D shape understanding and generation for real-world interiors.

Abstract

The 3D CAD shapes in current 3D benchmarks are mostly collected from online model repositories. Thus, they typically have insufficient geometric details and less informative textures, making them less attractive for comprehensive and subtle research in areas such as high-quality 3D mesh and texture recovery. This paper presents 3D Furniture shape with TextURE (3D-FUTURE): a richly-annotated and large-scale repository of 3D furniture shapes in the household scenario. At the time of this technical report, 3D-FUTURE contains 20,240 clean and realistic synthetic images of 5,000 different rooms. There are 9,992 unique detailed 3D instances of furniture with high-resolution textures. Experienced designers developed the room scenes, and the 3D CAD shapes in the scene are used for industrial production. Given the well-organized 3D-FUTURE, we provide baseline experiments on several widely studied tasks, such as joint 2D instance segmentation and 3D object pose estimation, image-based 3D shape retrieval, 3D object reconstruction from a single image, and texture recovery for 3D shapes, to facilitate related future researches on our database.

Figures (21)

• Figure 1: 3D-FUTURE. Top: Exquisite interior designs obtained from Alibaba Topping Homestyler design platform. Bottom: An overview of the properties of 3D-FUTURE. All the interior designs are developed or reviewed by experienced designers to ensure their quality. The photo-realistic synthetic scenes are rendered by the advanced rendering engine V-ray. The statistics of 3D-FUTURE are presented in Sec. \ref{['sec:properties_statistics']}.
• Figure 2: DFSM. An illustration of the deep furnishing suit model (DFSM) for deep visual embedding in Sec. \ref{['subsubsec:deep_visual_embedding']}. The development of the framework borrows the concepts from Bert devlin2018bert. We construct two tasks here, including mask prediction and compatibility scoring, as explained in Sec. \ref{['subsubsec:deep_visual_embedding']}. There is only one visual embedding network (VEN) which is shared in both the two tasks. The deep visual embedding ("orange") for a specific item is captured by the trained VEN.
• Figure 3: Realistic Renderings of Aesthetic Interior Designs. Left: experienced design templates. Right: created aesthetic interior designs. These AI generated designs are reviewed by designers. Zoom in for better view.
• Figure 4: 2D-3D Alignments. We provide precise 6DoF pose annotations for most of furniture shapes involved in each scene. zoom in for better view.
• Figure 5: Samples of the high-quality 3D shapes and their informative textures in 3D-FUTURE.