A Large Dataset of Object Scans

By Sungjoon Choi, Qian-Yi Zhou, Stephen D. Miller, Vladlen Koltun. 2016

246 citations recorded by the source corpus.

We have created a dataset of more than ten thousand 3D scans of real objects. To create the dataset, we recruited 70 operators, equipped them with consumer-grade mobile 3D scanning setups, and paid them to scan objects in their environments. The operators scanned objects of their choosing, outside the laboratory and without direct supervision by computer vision professionals. The result is a large and diverse collection of object scans: from shoes, mugs, and toys to grand pianos, construction vehicles, and large outdoor sculptures. We worked with an attorney to ensure that data acquisition did not violate privacy constraints. The acquired data was placed irrevocably in the public domain and is available freely at http://redwood-data.org/3dscan.

Subfields: Aerospace Engineering, Environmental Engineering, Geology, Evaluation & Benchmarks, 3D Vision & Pose

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Environmental Engineering

In Geology

In Evaluation & Benchmarks

In 3D Vision & Pose

Citations in the corpus

Revisiting Point Cloud Classification: A New Benchmark Dataset and Classification Model on Real-World Data (2019, background): “[8] proposed a dataset of more than 10,000 object scans in the real world.”
ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes (2017, methodology): “The median reconstruction processing time (including data conversion, dense voxel fusion, surface mesh extraction, alignment, cleanup, and preview thumbnail image rendering) is 11 .”
OctNet: Learning Deep 3D Representations at High Resolutions (2016, background): “At the same time, large 3D repositories such as ModelNet [47], ShapeNet [6] or 3D Warehouse1 as well as databases of 3D object scans [7] are becoming increasingly available.”
Pix3D: Dataset and Methods for Single-Image 3D Shape Modeling (2018, background): “[8], which contains a rich and diverse set of shapes, each with an RGB-D video.”
Shape Completion Using 3D-Encoder-Predictor CNNs and Shape Synthesis (2016, background): “Another option would be to generate training data from real-world input, captured with careful scanning and complete scanning patterns; e.g., using the dataset captured by Choi et al. [5].”
3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction (2016, background): “This trend has been boosted now that 3D printing is a democratized technology and 3D acquisition methods are accurate and eﬃ-cient [2].”
Matterport3D: Learning from RGB-D Data in Indoor Environments (2017, background): “Existing work on curation of RGB-D datasets has focused mostly on scans of individual objects [5], stand-alone rooms [20, 28], views of a room [32, 35], spaces from academic buildings and small apartments [7], and small collections of rooms or public spaces [44, 1, 22].”
Learning 3D Object Categories by Looking Around Them (2017, methodology): “-34 -34.27 e R on LDOS).”
DIST: Rendering Deep Implicit Signed Distance Function With Differentiable Sphere Tracing (2019, background): “Comparison on 3D shape prediction from multi-view images on real-world dataset [5].”
Deep Marching Cubes: Learning Explicit Surface Representations (2018, background): “However, the recent success of deep learning [19,20,38] and the availability of large 3D datasets [5, 6, 9, 26, 37] nourishes hope for models that are able to learn powerful 3D shape representations from data, allowing reconstruction even in the presence of missing, noisy and incomplete observations”
ABO: Dataset and Benchmarks for Real-World 3D Object Understanding (2021, background): “There exists a handful of publicly available datasets with large collections of 3D objects [5,6,14], but they don’t contain physically-accurate reflectance parameters that can be used for physically based rendering to generate photoreal-”
Thingi10K: A Dataset of 10, 000 3D-Printing Models (2016, methodology): “released a dataset of 10,000 scanned objects, with a subset of 383 successfully reconstructed 3D models [Choi et al. 2016].”
UnrealCV: Connecting Computer Vision to Unreal Engine (2016, background): “Computer vision researchers have created large 3D repositories and virtual scenes[4,10,15,6].”
ShapeFormer: Transformer-based Shape Completion via Sparse Representation (2022, methodology): “Then we show our approach can effectively handle a variety of shape types, outof-distribution shapes, and real-world scans from the Redwood dataset [16] (Sec.”
Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction (2021, background): “Choi et al. [9] provide a large set of realobject-centric RGB-D videos, however only a small subset is annotated 3D reconstruction A vast amount of methods studied fully supervised 3D reconstruction from 2D images making use of several different representations: voxel grids [10, 18], meshes [19,…”
3D Shape Segmentation with Projective Convolutional Networks (2016, unknown): “Chaudhuri acknowledges support from Adobe.”
LION: Latent Point Diffusion Models for 3D Shape Generation (2022, background): “[137] Sungjoon Choi, Qian-Yi Zhou, Stephen Miller, and Vladlen Koltun.”
Objaverse: A Universe of Annotated 3D Objects (2022, background): “Datasets of scanned 3D objects and environments are valuable for real-world understanding [10, 13, 14, 38], but are quite small and limited.”