Skip to content

Dataset · zeyuanyin

Multi-view renders of 83K Objaverse 3D objects

zeyuanyin/Dome-Objaverse

A set of 48 views for each of 83,296 3D objects for training 3D generation models.

What's inside

The dataset contains multi-view renders of 83,296 objects from the Objaverse collection, captured using a virtual camera dome. Exactly 48 images are prepared for each object: three rings of 12 views at elevation angles of 0°, 30°, and 60°, plus a top ring at 90°. Ready-made renders save researchers from resource-intensive 3D scene rendering from scratch.

What you can do with it

  • Training single- or multi-image 3D object reconstruction models (image-to-3d)
  • Training generative neural networks for text-to-3d asset creation
  • Training novel view synthesis models
Tasks
Image → 3d, Text → 3d
Size
10K–100K rows
Downloads this month
120.1K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("zeyuanyin/Dome-Objaverse")