Skip to content

Dataset · ShapeNet

3D objects in 3D Gaussian Splatting format

ShapeNet/ShapeSplatsV1

A collection of 52,000 3D objects as Gaussian splats from ShapeNetCore for 3D generation tasks.

What's inside

ShapeSplatsV1 features 52,000 3D objects from 55 categories of the ShapeNetCore dataset (the entire ShapeSplats project covers 65,000 objects from 87 categories collected from ShapeNetCore, ShapeNet-Part, and ModelNet). Data is distributed as .ply files, where parameters of each Gaussian are written in custom vertex attributes.

What you can do with it

  • Train generative models for creating and reconstructing 3D objects in Gaussian Splatting representation
  • Train neural networks for 3D scene classification and segmentation directly from Gaussian splats
Languages
English
Size
10K–100K rows
Downloads this month
557
Likes
48
Access
on request on Hugging Face

How to load it

from datasets import load_dataset

ds = load_dataset("ShapeNet/ShapeSplatsV1")