University of Toronto HSI-301 Dataset

The University of Toronto Hyperspectral-301 (UT-HSI-301) dataset consists of three hyperspectral images with 301 spectral channels, covering suburban, urban and forest landscapes in southern Ontario. The images were collected and pre-processed by the RSSEM Lab.

The dataset was used to study how spectral dimensionality reduction affects hyperspectral pixel classification: five methods (PCA, KPCA, ICA, autoencoders and denoising autoencoders) compressed the 301-dimensional pixels before classification. It is hosted by the Visual Computing Lab at Ontario Tech University.

This dataset can serve as an open benchmark for hyperspectral image analysis. With 301 spectral channels and reference polygons for suburban, urban and forest scenes, it can be used to develop and compare dimensionality reduction and classification methods, including machine learning and deep learning models, for mapping rooftops, roads, lawns and trees, as well as to design efficient processing pipelines for settings with limited computing resources.