Multimodal Dataset for Cinnamon Powder Adulteration Detection Using Colorimetry, FTIR Spectroscopy, and RGB Images

This dataset provides a comprehensive multimodal benchmark for the detection and quantification of cinnamon powder adulteration using three complementary sensing modalities: colorimetric measurements, Fourier Transform Infrared (FTIR) spectroscopy, and RGB images. The dataset was developed to support research in food authentication, adulteration detection, spectroscopy, computer vision, multimodal machine learning, and explainable artificial intelligence. The dataset comprises 5,850 verified data records, including 1,950 colorimetric measurements, 1,950 FTIR spectra, and 1,950 RGB images, collected from cinnamon powder adulterated with three economically important adulterants: peanut shell powder, walnut shell powder, and wood dust powder. Samples were prepared at thirteen adulteration levels (0%, 1%, 2%, 3%, 4%, 6%, 8%, 10%, 12%, 15%, 18%, 21%, and 24% w/w), with 50 independently prepared samples per adulteration level for each adulterant. Consistent sample identifiers are maintained across all three modalities, enabling straightforward multimodal data fusion and cross-modal analysis. Prior to publication, the complete dataset underwent systematic quality verification, including validation of sample identifiers, adulteration labels, duplicate detection, folder organization, and sample completeness. All verification checks were successfully completed.本数据集提供了一套全面的多模态基准测试集，用于通过三种互补传感模态检测并定量分析肉桂粉掺假情况，三种传感模态分别为比色测量、傅里叶变换红外（Fourier Transform Infrared, FTIR）光谱法以及RGB图像。本数据集的构建旨在支撑食品真实性鉴定、掺假检测、光谱学、计算机视觉、多模态机器学习以及可解释人工智能领域的相关研究。 本数据集共包含5850条经过验证的数据记录，涵盖1950条比色测量数据、1950条FTIR光谱数据以及1950条RGB图像数据。所有数据均采集自掺入三种具有重要经济价值掺假物的肉桂粉样本，三种掺假物分别为花生壳粉、核桃壳粉与木粉。 样本按照13种掺假梯度制备，梯度分别为0%、1%、2%、3%、4%、6%、8%、10%、12%、15%、18%、21%、24%（质量分数，w/w）。针对每种掺假物，每个掺假梯度均设置50份独立制备的样本。三种传感模态下均采用统一的样本标识符，可直接支持多模态数据融合与跨模态分析。 在正式发布前，完整数据集已完成系统性质量验证，验证内容涵盖样本标识符校验、掺假标签核验、重复数据检测、文件夹结构整理以及样本完整性核查，所有验证环节均顺利通过。