| --- |
| license: cc-by-nc-nd-4.0 |
| task_categories: |
| - image-to-image |
| - image-feature-extraction |
| tags: |
| - hyperspectral |
| - spectral-imaging |
| - remote-sensing |
| - computer-vision |
| - natural-images |
| - icvl |
| - bgu |
| pretty_name: ICVL Hyperspectral Dataset (2016) |
| size_categories: |
| - n<1K |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: preview/** |
| --- |
| |
| <p align="center"> |
| <img src="https://icvl.cs.bgu.ac.il/assets/research/hyperspectral-imaging/Cover_Graphic-1024x515.png" alt="ICVL Hyperspectral Imaging" width="820"/> |
| </p> |
|
|
| <h1 align="center">ICVL Hyperspectral Dataset (2016)</h1> |
|
|
| <p align="center"> |
| <b>202 natural-scene hyperspectral images · 400–1000 nm · Specim PS Kappa DX4</b><br/> |
| <sub>Interdisciplinary Computational Vision Laboratory · Ben-Gurion University of the Negev</sub> |
| </p> |
|
|
| <p align="center"> |
| <a href="https://icvl.cs.bgu.ac.il/pages/researches/hyperspectral-imaging.html" target="_blank" rel="noopener noreferrer">🌐 Project page</a> · |
| <a href="#citation">📄 Citation</a> · |
| <a href="#license">⚖️ License</a> |
| </p> |
|
|
| --- |
|
|
| ## 🔭 Overview |
|
|
| The **ICVL Hyperspectral Dataset** is a collection of high-resolution hyperspectral images of natural scenes — urban landscapes, rural views, indoor environments, plants, everyday objects and more — released by the [Interdisciplinary Computational Vision Laboratory (ICVL)](https://icvl.cs.bgu.ac.il/) at Ben-Gurion University of the Negev. |
|
|
| Images were acquired with a **Specim PS Kappa DX4** hyperspectral camera mounted on a **rotary stage** for spatial (line) scanning, yielding dense spectral cubes across the visible and near-infrared range. |
|
|
| <p align="center"> |
| <img src="https://icvl.cs.bgu.ac.il/assets/research/hyperspectral-imaging/Screenshot_102017_030853_PM-1024x322.jpg" alt="Sample hyperspectral scenes" width="820"/> |
| </p> |
|
|
| ## 📊 Specifications |
|
|
| | Property | Raw data (ENVI) | MAT (downsampled) | |
| |---|---|---| |
| | **Spectral range** | 400 – 1000 nm | 400 – 700 nm | |
| | **Spectral bands** | 519 (~1.25 nm step) | 31 (10 nm step) | |
| | **Spatial resolution** | 1392 × 1300 | 1392 × 1300 | |
| | **Format** | `.raw` + `.hdr` | `.mat` (HDF5) | |
| | **Scenes** | 202 | 202 | |
|
|
| ## 📁 Repository structure |
|
|
| ``` |
| ICVL_HS_2016/ |
| ├── mat/ # 202 files — downsampled 31-band cubes (400–700 nm) |
| │ └── <scene>.mat |
| ├── raw/ # 404 files — full 519-band ENVI cubes (400–1000 nm) |
| │ ├── <scene>.raw |
| │ └── <scene>.hdr |
| ├── preview/ # 202 files — RGB JPEG previews |
| │ └── <scene>.jpg |
| ├── file_list.txt # optional splits / subset list |
| └── README.md |
| ``` |
|
|
| Total: **809 data files · ~335 GB**. |
|
|
| Scene stems (e.g. `4cam_0411-1640`, `bguCAMP_0514-1711`) are shared across `mat/`, `raw/` and `preview/`, so each scene can be referenced by its stem. |
|
|
| ## 🖼️ Sample scenes |
|
|
| <p align="center"> |
| <img src="preview/plt_0411-1046.jpg" width="240"/> |
| <img src="preview/bgu_0403-1459.jpg" width="240"/> |
| <img src="preview/plt_0411-1116.jpg" width="240"/><br/> |
| <img src="preview/sat_0406-1129.jpg" width="240"/> |
| <img src="preview/rsh_0406-1441-1.jpg" width="240"/> |
| <img src="preview/bgu_0403-1525.jpg" width="240"/> |
| </p> |
|
|
| <p align="center"><sub>RGB previews of hyperspectral cubes — plants, urban scenes, indoor objects.</sub></p> |
|
|
| ## 🚀 Loading the data |
|
|
| ### Downsampled `.mat` cubes (recommended for most vision tasks) |
|
|
| MATLAB v7.3 (HDF5-backed). Each file contains: |
|
|
| - **`rad`** — hyperspectral radiance cube, shape `(1300, 1392, 31)`, wavelengths 400 – 700 nm at 10 nm steps. |
| - **`bands`** — vector of 31 wavelength values (nm). |
|
|
| ```python |
| import h5py, numpy as np |
| with h5py.File("mat/<scene>.mat", "r") as f: |
| cube = np.array(f["rad"]) # transpose to (H, W, C) as needed |
| bands = np.array(f["bands"]) |
| ``` |
|
|
| ### Full-range ENVI `.raw` + `.hdr` cubes |
|
|
| 519 spectral bands over 400 – 1000 nm. Read with the [`spectral`](https://www.spectralpython.net/) Python package or any ENVI-aware tool: |
|
|
| ```python |
| import spectral |
| img = spectral.envi.open("raw/<scene>.hdr", "raw/<scene>.raw") |
| cube = img.load() # (1300, 1392, 519) |
| ``` |
|
|
| ### RGB previews |
|
|
| Standard sRGB JPEGs for quick browsing and dataset navigation — no radiometric use. |
|
|
| ## 🎯 Intended uses |
|
|
| - **Spectral super-resolution** — recovering hyperspectral signals from RGB inputs |
| - **Hyperspectral denoising, super-resolution, compression** |
| - **Spectral unmixing** and material identification |
| - **Benchmarking** spectral–spatial deep-learning models (NTIRE spectral-recovery challenges have used this dataset extensively) |
|
|
| <a id="license"></a> |
| ## ⚖️ License |
|
|
| Released under **[CC BY-NC-ND 4.0](https://creativecommons.org/licenses/by-nc-nd/4.0/)** — Attribution · NonCommercial · NoDerivatives. |
|
|
| You may **share** the dataset for non-commercial purposes with attribution to the original authors; derivative distributions are not permitted. |
|
|
| <a id="citation"></a> |
| ## 📄 Citation |
|
|
| If you use this dataset in academic work, please cite: |
|
|
| ```bibtex |
| @inproceedings{arad_and_ben_shahar_2016_ECCV, |
| title = {Sparse Recovery of Hyperspectral Signal from Natural RGB Images}, |
| author = {Arad, Boaz and Ben-Shahar, Ohad}, |
| booktitle = {European Conference on Computer Vision (ECCV)}, |
| pages = {19--34}, |
| year = {2016}, |
| organization = {Springer} |
| } |
| ``` |
|
|
| ## 📮 Contact |
|
|
| Interdisciplinary Computational Vision Laboratory (ICVL) |
| Ben-Gurion University of the Negev |
| 🔗 <a href="https://icvl.cs.bgu.ac.il/pages/researches/hyperspectral-imaging.html" target="_blank" rel="noopener noreferrer">icvl.cs.bgu.ac.il — Hyperspectral Imaging</a> |
|
|
| **Correspondence:** Prof. Ohad Ben-Shahar (PI) — [ben-shahar@cs.bgu.ac.il](mailto:ben-shahar@cs.bgu.ac.il) |
|
|
