Skip to content

Dataset · desmondlzy

Synthetic scenes with diverse lighting for 3D Gaussian Splatting

desmondlzy/bigs-data

7 synthetic 3D scenes with complex lighting effects and dozens of viewpoints for relighting tasks.

What's inside

The dataset consists of 7 synthetic scenes ranging from 1.8 to 3.2 GB each featuring complex optical effects: subsurface scattering, fuzzy surfaces, and rainbow reflections. Images are prepared in OLAT (one-light-at-a-time) format for each scene: 40 training conditions with 48 viewpoints and 58 test conditions. The dataset was created for a 3DV 2025 conference research paper on relightable 3D Gaussian Splatting.

What you can do with it

  • Training and testing relighting models and 3D Gaussian Splatting
  • Evaluating rendering quality of complex materials and lighting effects in neural graphics
Downloads this month
84.7K
Likes
1

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("desmondlzy/bigs-data")