SkinSim
A statistical project for human skin reflectance spectrum. The reflectance spectrum of human skin is a crucial aspect of dermatological research, cosmetic science, and medical diagnostics. This project aims to analyze and model the reflectance properties of human skin across different wavelengths of light.
1 Datasets
The project will utilize datasets that include measurements of skin reflectance across various wavelengths. These datasets include real measurements sourced from public repositories, and synthetic data simulated through biophysical models.
- Real measurements: International Skin Spectra Archive (ISSA) (Lu et al. 2025), a multicultural human skin phenotype dataset, containing 15,256 records of both spectral and colorimetric data derived from 2,113 subjects. This dataset covers a wide variability in skin characteristics such as geography, ethnicity, age, gender, and body location.
- Synthetic data: Generated through a biophysical layered skin model and Monte Carlo light transport simulations (Aliaga et al. 2023). This approach learns the mapping between the skin parameters and diffuse skin reflectance through an encoder-decoder network.
2 Statistical analyses
The BioSkin validation results summarize the analysis of 4,000 generated spectra: 1,167 pass MI–EI screening and 733 (18.325%) pass both validation stages. See the executed BVCM notebook for the full analysis and the rerun report for the paired comparison and audit.
The project will perform various statistical analyses to explore the properties of skin reflectance. These analyses include:
- Descriptive statistics to summarize the reflectance data.
- Reflectance spectrum filter to classify realistic skin reflectance spectra from synthetic ones.
- Shape analysis to explore the relationship between skin properties and reflectance shapes.
- Bayesian varying coefficient models to investigate the effects of skin parameters on skin reflectance.