Datasets:

hyper3labs
/

amazon-berkeley-objects

Amazon Berkeley Objects is a collection of product listings with multilingual metadata, catalog imagery, high-quality 3d models with materials and parts, and benchmarks derived from that data.

License

This work is licensed under the Creative Commons Attribution 4.0 International Public License. To obtain a copy of the full license, see LICENSE-CC-BY-4.0.txt, visit CreativeCommons.org or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.

Under the following terms:

Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.
No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.

Attribution

Credit for the data, including all images and 3d models, must be given to:

Amazon.com

Credit for building the dataset, archives and benchmark sets must be given to:

Matthieu Guillaumin (Amazon.com), Thomas Dideriksen (Amazon.com), Kenan Deng (Amazon.com), Himanshu Arora (Amazon.com), Jasmine Collins (UC Berkeley) and Jitendra Malik (UC Berkeley)

Description

Amazon Berkeley Objects is a collection of 147,702 product listings with multilingual metadata and 398,212 unique catalog images. 8,222 listings come with turntable photography (also referred as spin or 360º-View images), as sequences of 24 or 72 images, for a total of 586,584 images in 8,209 unique sequences. For 7,953 products, the collection also provides high-quality 3d models, as glTF 2.0 files.

The collection is made of the following directories and files:

README.md - The present file.
LICENSE-CC-BY-4.0.txt - The License file. You must read, agree and comply to the License before using the Amazon Berkeley Objects data.
listings/ - Product description and metadata. Check listings/README.md for details. archives/abo-listings.tar contains all the files in listings/ as a tar archive.
images/ - Catalog imagery, in original and smaller (256px) resolution. Check images/README.md for details. archives/abo-images-original.tar contains the metadata and original images from images/original/ as a tar archive and archives/abo-images-small.tar contains the metadata and downscaled images from images/small/ as a tar archive.
spins/ - Spin / 360º-View images and metadata. Check spins/README.md for details. archives/abo-spins.tar contains the metadata and images from spins/ as a tar archive.
3dmodels/ - 3D models and metadata. Check 3dmodels/README.md for details. archives/abo-3dmodels.tar contains the metadata and 3d models from 3dmodels/ as a tar archive.
benchmarks/abo-mvr.csv.xz - Train/val/test dataset splits for the Multi- View Retrieval experiments of the CVPR 2022 ABO paper
archives/abo-benchmark-material.tar - Train/test dataset for the Material Prediction experiments of the CVPR 2022 ABO paper. See the README.md file in the archive for more details.
archives/abo-part-labels.tar - Dataset for the 2023 ABO Fine-grained Semantic Segmentation Competition organized for the 3D Vision and Modeling Challenges in eCommerce Workshop in conjunction with ICCV 2023.

Footnotes

[^1]: Importantly, there is no guarantee that these URLs will remain unchanged and available on the long term, we thus recommend using the images provided in the archives instead.