Hyper-Skin

Hyper-Skin is a dataset containing the human facial skin image captured by hyperspectral camera. It covers a wide spectral range from visible (VIS) to near-infrared (NIR) spectra, allowing for a holistic understanding of various aspects of human facial skin. The dataset includes 306 hyperspectral cubes captured from 51 subjects (51 subjects * 6 images (neutral: {front, left, right}, smile: {front, left, right})), featuring diverse facial poses and angles. Each hyperspectral cube with spatial resolution of 1024 x 1024 contains 1,048,576 spectra of length 448, providing a rich and detailed representation of skin properties. The 448-bands data is resampled into two 31-band datasets: one covering the visible spectrum from 400nm to 700nm, and the other covering the near-infrared spectrum from 700nm to 1000nm. Additionally, synthetic RGB and Multispectral (MSI) data are generated, comprising RGB images and an infrared image at 960nm. These steps result in the creation of our Hyper-Skin dataset, encompassing two types of data: (RGB, VIS) and (MSI, NIR). Note that part of the dataset, specifically the dataset with synthetic RGB generated from CIE color function, was used in the ICASSP2024 Grand Challenge.

Hyper-Skin是一款由高光谱相机（hyperspectral camera）采集的人类面部皮肤图像数据集。该数据集覆盖从可见光（VIS）到近红外（NIR）的宽光谱范围，可实现对人类面部皮肤多维度特性的全面认知。数据集共包含来自51名受试者的306个高光谱立方体（hyperspectral cubes），具体为51名受试者×6张图像（中性表情：正面、左侧、右侧；微笑表情：正面、左侧、右侧），涵盖多样化的面部姿态与拍摄角度。每个空间分辨率为1024×1024的高光谱立方体包含1048576个长度为448的光谱采样点，能够提供丰富且细致的皮肤属性表征。该448波段的数据被重采样为两个31波段的数据集：其一覆盖400nm至700nm的可见光光谱范围，其二覆盖700nm至1000nm的近红外光谱范围。此外，本数据集还生成了合成RGB数据与多光谱（MSI）数据，其中多光谱数据包含RGB图像与一张960nm的红外图像。通过上述步骤最终构建完成Hyper-Skin数据集，其包含两类数据：（RGB、可见光）与（多光谱、近红外）。值得注意的是，该数据集的一部分——即基于CIE颜色函数（CIE color function）生成合成RGB的子集——曾被用于ICASSP2024大奖赛（ICASSP2024 Grand Challenge）。