Skoltech3D: Multi-sensor large-scale dataset for multi-view 3D reconstruction

Oleg Voynov	Gleb Bobrovskikh	Pavel Karpyshev	Saveliy Galochkin
Andrei-Timotei Ardelean	Arseniy Bozhenko	Ekaterina Karmanova	Pavel Kopanev
Yaroslav Labutin-Rymsho	Ruslan Rakhimov	Aleksandr Safin	Valerii Serpiva
Alexey Artemov	Evgeny Burnaev	Dzmitry Tsetserukou	Denis Zorin

Abstract

We present a multi-sensor dataset for multi-view 3D surface reconstruction. It includes registered RGB and depth data from 7 sensors of different resolutions and modalities (a): smartphones, Intel RealSense, Microsoft Kinect, industrial cameras, and structured-light scanner. The scenes are selected to emphasize a diverse set of material properties challenging for existing algorithms (c), such as featureless (F), highly specular with sharp reflections (S), or translucent (T), as illustrated with reconstructions produced by state-of-the-art algorithms (compare with an “easy” object on the bottom right). We provide around 1.4 million images of 107 different scenes acquired from 100 viewing directions under 14 lighting conditions (b). We expect our dataset will be useful for evaluation and training of 3D reconstruction algorithms and for related tasks.

Resources

Fulltext PDF: Skoltech3D: Multi-sensor large-scale dataset for multi-view 3D reconstruction
Supplementary PDF: Skoltech3D: Multi-sensor large-scale dataset for multi-view 3D reconstruction Supplementary Materials
Source code: https://github.com/Skoltech-3D/sk3d_data
Video Video
Official project page: https://skoltech3d.appliedai.tech

Citing:

@inproceedings{2023-sk3d,
 author = {Voynov, Oleg and Bobrovskikh, Gleb and Karpyshev, Pavel and Galochkin, Saveliy and Ardelean, Andrei-Timotei and Bozhenko, Arseniy and Karmanova, Ekaterina and Kopanev, Pavel and Labutin-Rymsho, Yaroslav and Rakhimov, Ruslan and others},
 booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
 pages = {21392--21403},
 title = {Multi-sensor large-scale dataset for multi-view 3D reconstruction},
 year = {2023}
}

Skoltech3D: Multi-sensor large-scale dataset for multi-view 3D reconstruction

Abstract

Video

Resources