Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
The dataset viewer is not available for this split.
Server error while post-processing the rows. Please report the issue.
Error code:   RowsPostProcessingError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

ABO Furniture — web-optimised GLB

Web-weight 3D models of real retail furniture products, derived from Amazon Berkeley Objects (ABO).

Every model is a real product listing: it keeps its ASIN, so glb/<ASIN>.glb corresponds to https://www.amazon.com/dp/<ASIN>, and the listing's brand, product name and stated dimensions travel with it.

Why this exists

ABO's original models are artist-made and beautiful, but they are built for offline rendering: the full set is 154 GB, roughly 20 MB per model, with 4096px textures. That is unusable in a browser on a phone.

This derivative re-encodes each one for the web. A representative example — a tufted velvet sofa — goes from 42.45 MB to 356 KB, a 119x reduction, with no visible difference at room-planner viewing distance.

What was changed

Each model was processed with glTF-Transform:

  • textures resized to 1024px and re-encoded as WebP (EXT_texture_webp)
  • geometry welded, joined, de-duplicated and pruned
  • EXT_meshopt_compression + KHR_mesh_quantization

Note that EXT_meshopt_compression requires a decoder. three.js and <model-viewer> differ here: three/drei handle it once MeshoptDecoder is registered, while <model-viewer> will fail to load these files.

Contents

  • glb/<ASIN>.glb — the optimised model
  • thumbs/<ASIN>.webp — the product's own catalog photograph, 256px
  • index.json — brand, product name, category, model extents in metres, listed product dimensions, file size

Licence and attribution

CC BY 4.0, inherited from ABO. (ABO was relicensed from CC BY-NC to CC BY in 2023; older write-ups still say NC.)

Credit for the data, including all images and 3D models:

Amazon.com

Credit for building the original dataset:

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)

Changes made: geometry and texture optimisation for web delivery, as described above. No geometry was simplified and no model was otherwise altered.

Downloads last month
1,003