Published September 6, 2026 | Version v1

KPI-NeRF Real-World and Synthetic Scene

  • 1. ROR icon University of Siegen
  • 2. ROR icon Nanjing University

Description

KPI-NeRF Real-World and Synthetic Scene

This folder contains the scenes used in the KPI-NeRF experiments for spectral and plenoptic light-field neural rendering. It is organized by scene name so that each scene can be downloaded and used independently.

Contents

  • Bulldozer, Dragon, and Ajar: synthetic EXR scenes with camera transforms and ground-truth data.
  • Ball, Bird, ColorChecker, and Wild Bush: real-world reconstructed spectral scenes with multiple views.
  • Lucky Cat: reconstructed image sequences from DSC_0002 through DSC_0013.
  • Flower: reconstructed image sequences from DSC_0001 through DSC_0013.

The original scene layouts and file formats are preserved inside each scene directory. EXR files contain the multi-channel recorded or rendered views; JSON files contain camera and scene metadata where available; NumPy files contain auxiliary ground-truth or illumination data where provided.

Directory layout

KPI-NeRF_Scenes/
├── Ajar/
├── Ball/
├── Bird/
├── Bulldozer/
├── ColorChecker/
├── Dragon/
├── Flower/
├── Lucky Cat/
└── Wild Bush/

The dataset is intended for research and benchmarking in neural radiance fields, spectral rendering, computational photography, and light-field imaging. Please cite the associated KPI-NeRF publication and this Zenodo record when using the data.

Files

KPI-NeRF_Scenes.zip

Files (1.8 GB)

Name Size
md5:4a4541ef9cb887d2c0603fd19dd1fbdc
1.8 GB Preview Download

Additional details

References

  • KPI-NeRF: Hyperspectral Neural Radiance Fields from a Single Kaleidoscopic Plenoptic Image Erqi Huang, John Restrepo, Xun Cao, Ivo Ihrke Computer Graphics Forum, 2026