A HS Dataset Collected at Strathclyde University

Note: This dataset was captured by UG EEE students at the University of Strathclyde. 

Overview

This dataset contains 26 uncalibrated hyperspectral images captured at Strathclyde University. Each image is provided with hyperspectral and annotation data. Seven material classes have been used to label the dataset: artificial, rock, sky, soil, water, wood, vegetation.

For each sample, the following files are included:

image.npy – a NumPy array containing the hyperspectral data with shape (800, 800, 600).

image.json – annotations created using LabelMe, describing object masks and regions of interest.

baslerPIA1600_calibration.txt – a text file containing the wavelengths corresponding to each of the 600 spectral bands captured by the camera. These values are crucial for interpreting the spectral dimension of the data.

Convert JSON to Annotated Image

You can convert the `.json` annotations into masks and visualisations using the LabelMe tools. 

Install labelme if you haven't:pip install labelme

Run the following command inside a sample directory:

    labelme_json_to_dataset image.json

This will generate a folder image_json/ containing:

label.png – a PNG mask of labels

label_viz.png – a visual representation of the annotation

label_names.txt – list of class labels