Spectral reflectance dataset of soybean varieties measured using Vis/NIR spectroscopy with an AS7265X multispectral sensor for quality classification

This dataset contains multispectral reflectance measurements of soybean samples collected using a portable Visible/Near-Infrared (Vis/NIR) spectroscopy system equipped with an AS7265X multispectral sensor. The sensor records spectral responses across 18 discrete wavelength channels ranging from 410 nm to 940 nm, covering the ultraviolet (UV), visible (VIS), and near-infrared (NIR) spectral regions.

The dataset was developed to support the analysis and classification of soybean quality using non-destructive spectroscopy combined with machine learning techniques. Each sample entry includes spectral reflectance intensity values from the AS7265X sensor along with categorical labels describing the soybean variety group and quality class.

The soybean samples used in this dataset are non-GMO Indonesian soybean varieties, consisting of Anjasmoro and Grobogan. To represent both seed and consumption categories, the samples are grouped into four variety labels:
- BA (Anjasmoro soybean seed)
- BG (Grobogan soybean seed)
- KA (Anjasmoro soybean for consumption)
- KG (Grobogan soybean for consumption)

In addition to the variety label, each sample is assigned a quality class representing soybean quality levels:
- Poor
- Fair
- Premium

Each dataset record contains 18 spectral reflectance intensity values, one Varietas column indicating the soybean group (BA, BG, KA, KG), and one Class column representing the quality category (Poor, Fair, Premium). The spectral values represent the original sensor outputs in arbitrary units (a.u.), enabling flexible preprocessing and modeling for spectral analysis and machine learning applications.

This dataset can support research in agricultural engineering, seed quality assessment, spectral feature analysis, and machine learning-based classification of soybean quality. It also provides a basis for developing portable spectroscopy systems for rapid and non-destructive evaluation of soybean seeds and consumption-grade soybeans.

The data collection activities were conducted in 2025. This research was financially supported by the Directorate of Research and Community Service, Directorate General of Research and Development, Ministry of Higher Education, Science, and Technology, under Contract No. 125/C3/DT.05.00/PL/2025.

Key Features:
- Multispectral reflectance data from 18 wavelength channels (410–940 nm).
- Non-destructive measurement using portable Vis/NIR spectroscopy.
- Includes two Indonesian soybean varieties: Anjasmoro and Grobogan.
- Four soybean sample groups representing seed and consumption categories.
- Three soybean quality classes: Poor, Fair, and Premium.
- Suitable for machine learning classification, seed quality prediction, and spectral analysis.