Dataset · CaiYuanhao
HDR scenes for novel view synthesis (Gaussian Splatting)
CaiYuanhao/HDR-GS
High dynamic range (HDR) image dataset for training and testing 3D scene rendering models with exposure control.
What's inside
Contains recalibrated and reorganized image scenes from the official HDR-GS project repository (NeurIPS 2024 paper). Featuring between 1,000 and 10,000 examples, the represented scenes include a bathroom, chair, dog, bear, table, and Sponza. The data is intended for Novel View Synthesis tasks with high dynamic range and exposure management.
What you can do with it
- Training and testing 3D Gaussian Splatting models with HDR support
- Comparing scene reconstruction algorithms under varying exposures
- Size
- 1K–10K rows
- Downloads this month
- 637
- Likes
- 2
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("CaiYuanhao/HDR-GS")