Hyperspectral Unmixing dataset created from hyperspectral data gathered usign SPECIM push-broom hyperspectral camera mounted on a UAV flying over blueberry fields in Lithuania. Created dataset contains six data classes and linear mixtures from raw data. All data is given in Python Numpy array .npy files. To keep the annonimity of data owners only the non georectified data cubes are published. Dataset includes three hyperpectral data cubes of blueberry fields and Dark reference cube to show camera noise. Data structure: cube_1, cube_2, cube_2 and Dark - folder with hyperspectral data. calibration_data.npy - Data of calibration plates (with 40%, 10% and 5% reflectance values) from hyperspectral flight that were used to conver DN to reflectance. endmembers.npy - Spectra of siz endmembers (classes) used in the dataset. cube_x folders include: class_matrix.npy - Numpy matrix file of hyperspectral image classes (classification results) raw_data.npy - Hyperspectral cube created from raw camera data (with DN values) data_cube_3_0.npy and abundances_3_0.npy - Classified and mixed (using slidin window of 3x3 pixels with no overlap) hyperspectral data cube and class abundance matrix. endmember_errors.npy - matrix of variation for each of endmembers in the hyperspectral cube (used for evaluation mostly.) Dark folder: includes data folder with raw-dark_fl1_20230830_140006_radiance.dat and .hdr ENVI raster data files (library like rasterio for Python can used to read these files). This is the dark (0% reflectance) data cube and header file used in calibration.