SPEC-2B Dataset: a Visible-Near Infrared (VIS-NIR) Reflectance Spectra of Blueberry Fruits Across Cultivars and Maturity Stages

Reliable assessment of blueberry fruit maturity using optical proximal sensing requires standardized and well-documented spectral datasets covering multiple cultivars and ripening stages. To this end, we present the SPEC-2B Dataset, a curated collection of visible-near infrared (VIS-NIR) reflectance spectra acquired from blueberry (Vaccinium corymbosum L.) fruits representing three commercial cultivars (Duke, Legacy, and Blue Ribbon) and three visually assessed maturity categories. The dataset comprises 9588 reflectance spectra acquired using a compact spectrometer under consistent illumination geometry and acquisition protocols. Following preprocessing and quality control, a working spectral range from approximately 500 to 1000 nm was retained for analysis. All spectra were processed using a documented preprocessing workflow that includes dark signal correction, white reference normalization, integration-time normalization, wavelength trimming, and quality control procedures, including an empirical O2-A band index and an optional filtered subset to reduce localized spectral perturbations around 760 nm. Technical validation was conducted using descriptive statistics, empirical signal-to-noise ratio characterization, and principal component analysis (PCA) to examine internal consistency, spectral stability, and variance structure across cultivars and maturity categories. The dataset, associated metadata, raw spectral measurements, quality-control files, and visualization scripts are made publicly available to support reproducible research in agricultural spectroscopy, proximal sensing, and fruit quality assessment.