ICVL: Hyperspectral Image Database with 519 Spectral Bands

ICVL is a hyperspectral image dataset collected for the research 'Sparse Recovery of Hyperspectral Signal from Natural RGB Images'. It contains at least 200 images captured with a Specim PS Kappa DX4 camera and a rotary stage, with a spatial resolution of 1392 x 1300 pixels. The dataset was last updated on October 1, 2024, by the author danaroth.

Use Cases

• Developing sparse recovery models for hyperspectral signals based on the dataset's stated research purpose.
• Training spectral super-resolution algorithms based on the 519 spectral bands covering 400-1000nm.
• Benchmarking hyperspectral image processing techniques based on the high spatial resolution (1392 x 1300).
• Simulating and analyzing natural scene reflectance based on the dataset's spectral range and image content.

Strengths

• High spectral resolution with 519 bands covering 400-1000nm at roughly 1.25nm increments.
• High spatial resolution of 1392 x 1300 pixels per image.
• Dataset is actively growing, with at least 200 images currently available.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

danaroth on Hugging Face

Collection Method

Images acquired using a Specim PS Kappa DX4 hyperspectral camera and a rotary stage for spatial scanning.

Time Range

null

Freshness

Last updated 2024-10-01 15:26:14; freshness should be verified.

Geography

null