HyperPistachio: A Hyperspectral Image Dataset of Aflatoxin B1 Contaminated Pistachio Nuts

HyperPistachio: A Dataset for Aflatoxin Detection from Hyperspectral ImagesThis dataset contains hyperspectral images of raw pistachio samples contaminated with various levels of Aflatoxin in laboratory conditions. This collection aims to provide baseline data for the research and development of machine learning and image processing algorithms for non-destructive detection and quantification of Aflatoxin contamination in pistachios. The dataset consists of three images for each of the twenty-two levels of aflatoxin contamination, each containing 4,000 images. The data is organized into 22 sub-folders. For each image sample, three files are provided: 1. A binary data file in .bil (Band Interleaved by Line) format contains the hyperspectral data cube. A header file in .hdr format, containing metadata and technical information about the image. A preview image file in .png format for quick visual inspection of the sample. Technical Specifications: Number of Spectral Bands: 462 Wavelength Range: 386.88 nm to 1003.60 nm 2. Contamination Data and Levels (Ground Truth)The table below details the precise Aflatoxin contamination level for each sample. These values serve as the ground truth for training and evaluating models. The concentration unit is µg/kg (micrograms per kilogram), which is equivalent to ppb (parts per billion). Sample Name Aflatoxin Concentration (µg/kg)Level 00 0.00Level 01 0.40Level 02 0.67Level 03 0.88Level 04 1.13Level 05 1.66Level 06 2.48Level 07 2.15Level 08 2.30Level 09 3.05Level 10 2.82Level 11 3.01Level 12 3.85Level 13 4.43Level 14 5.30Level 15 5.12Level 16 8.93Level 17 6.37Level 18 26.14Level 19 33.17Level 19 33.17Level 20 TBALevel 21 TBA Export to Sheets3. Sample Code for Reading and VisualizationA sample script is provided to help you get started with the data. This code offers the following functionalities: Reading .bil and .hdr files. Displaying the image in pseudo-RGB and grayscale. Plotting the spectral profile for a selected pixel. How to Use:Please run the Visualise.m file using MATLAB version R2018b or later. Before running the script, ensure you have installed the required Toolboxes, such as Image Processing Toolbox and Mapping Toolbox.