Dataset Description

This is a database of fifty indoor‑outdoor scenes captured under daylight and an additional twenty‑five indoor scenes captured under artificial and mixed illumination. The images were acquired with a commercial hyperspectral camera (Nuance FX, CRI Inc.) equipped with an integrated liquid‑crystal tunable filter, which can sequentially adjust the filter to capture thirty‑one narrow spectral bands (each about 10 nm wide, centered from 420 nm to 720 nm in 10 nm steps).

The camera is equipped with an achromatic lens, and images are captured at the smallest feasible aperture to largely avoid chromatic aberration. All images are of static scenes and include masks to hide regions that moved during exposure.

The dataset is split into two parts:

• CZ_hsdb: 50 daylight indoor‑outdoor images (5.3 GB)
• CZ_hsdbi: 27 artificial/mixed‑light indoor images (2.2 GB)

It is provided for non‑commercial research use only. See the README.txt in each archive for details.

Dataset Features

This real‑world hyperspectral image database is available for non‑commercial research. It includes:

• CZ_hsdb: 50 daylight indoor‑outdoor images (5.3 GB)
• CZ_hsdbi: 27 artificial/mixed‑light indoor images (2.2 GB)

Citation

If you use this data in academic publications, please cite the following paper:

Ayan Chakrabarti and Todd Zickler, "Statistics of Real‑World Hyperspectral Images," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2011.

@conference{chakrabarti2011statistics,
  title={{Statistics of Real-World Hyperspectral Images}},
  author={Chakrabarti, A. and Zickler, T.},
  booktitle={Proc.~IEEE Conf.~on Computer Vision and Pattern Recognition (CVPR)},
  pages={193--200},
  year={2011}
}