Published July 23, 2026 | Version v1

Relightable Gaussian Splatting for Virtual Production Using Image-Based Illumination

Description

This is a dataset for the paper titled "Relightable Gaussian Splatting for Virtual Production Using Image-Based Illumination".

  • The dataset is captured using 19 multi-view cameras and a Virtual Production stage (small and normal size stage).
  • The dataset is split into three scenes and contains a set of static objects of various materials and textures. This included but is not limited to wood, plastic, fur and glass.
  • Each frame captures the stage using a different image-based background. 99 Backgrounds were captured for each scene.
  • The dataset also provides the estimated camera poses, instrinsic settings and point cloud, predicted using Nerfstudio COLMAP.

Files

scene1.zip

Files (4.6 GB)

Name Size
md5:34e26a4107fe2e2e452c9e7b637fb32e
2.3 GB Preview Download
md5:d8408b95aaf056f8eac0a8d8d6858f1a
1.5 GB Preview Download
md5:a6a14035150a2f24ac11cc64e6dfd26e
692.7 MB Preview Download

Additional details

Related works

Is supplement to
Preprint: arXiv:2605.09024 (arXiv)

Funding

UK Research and Innovation
EPSRC DTP MyWorld Scholarship SIPF00006/1