3) sensAIfood cereal NIR data (wheat, maze, barley, spelt) and reference protein and moisture (CRA-W set)

Nir spectra dataset of barley, spelt, wheat and maze grains, with either protein, moisture or both reference measurements. Contributions from CRAW-W (Centre wallon de Recherches agronomiques) represented by Dr. Vincent Baeten. Different datasets were taken with different spectrometers: Foss XDS Monochromator XM-1000, range 400 - 2500 nm and Foss NIRSYSTEM-5000, range 1100 - 2500. These datasets consists in data from different years, from different cereals from Belgium. The nomenclature used for the file names has the following structure. Cereal_sensAIfood_Y_NSamples_Spectrometer_CRAW The data is distributed in 3 folders containing datasets acquired from different spectrometers and divided into Protein only, Moisture only and Protein and Moisture reference values. Please consult the corresponding _metadata.xls for details about the reference measurements. MOISTURE Barley_sensAIfood_Moist_337_XDS_CRAW.csv Barley_sensAIfood_Moist_985_NIRS5000_CRAW.csv Maize_sensAIfood_Moist_585_NIRS5000_CRAW.csv Spelt_sensAIfood_Moist_198_XDS_CRAW.csv Spelt_sensAIfood_Moist_411_XDS_CRAW.csv Wheat2_sensAIfood_Moist_500_XDS_CRAW.csv WheatDurum_sensAIfood_Moist_170_XDS_CRAW.csv Wheat_sensAIfood_Moist_3305_NIRS5000_CRAW.csv Wheat_sensAIfood_Moist_500_XDS_CRAW.csv PROTEIN AND MOISTURE Barley_sensAIfood_Prot_Moist_636_XDS_CRAW.csv WheatDurum_sensAIfood_Prot_Moist_500_XDS_CRAW.csv PROTEIN Barley_sensAIfood_Protein_2096_NIRS5000_CRAW.csv Maize_sensAIfood_Protein_549_NIRS5000_CRAW.csv Spelt_sensAIfood_Protein_411_XDS_CRAW.csv Spelt_sensAIfood_Protein_45_XDS_CRAW.csv WheatDurum_sensAIfood_Protein_125_XDS_CRAW.csv Wheat_sensAIfood_Protein_231_XDS_CRAW.csv Wheat_sensAIfood_Protein_500_XDS_CRAW.csv Wheat_sensAIfood_Protein_5046_NIRS5000_CRAW.csv We also include an example import script in python (.ipynb) NOTE: Please note that the wet chemistry lab reference measurements of protein in different datasets/cereals use different reference measurements (DM=6.25, DM=5.7, as-is, etc.). This impacts the overall scale of protein content and should be taken into account if you are mixing datasets (e.g. one for model training and another for validation). All the information about the reference methods is available in the accompanying metadata files.