Project Vision

This project aims to develop an innovative framework for hyperspectral skin analysis by leveraging the Hyper-Skin dataset, which spans from the visible (VIS) to the near-infrared (NIR) spectrum. Our goal is to enable accurate skin spectra reconstruction from RGB images, facilitating the study of skin characteristics like melanin and hemoglobin concentrations directly on consumer devices.

The Hyper-Skin dataset is composed of 306 hyperspectral cubes from 51 subjects, capturing facial skin from three different angles and two facial poses. Each hyperspectral cube has a dimension of 1024x1024x448, and has been resampled into two separate 31-band datasets, covering the VIS spectrum from 400nm to 700nm and the NIR spectrum from 700nm to 1000nm. This extensive dataset enables the study of both surface-level and deeper tissue properties of the skin, offering valuable insights for applications in cosmetology and personalized skincare.

By bridging the gap between hyperspectral imaging and accessible technology, we seek to advance research in cosmetology and personalized skincare, fostering new applications that enhance skin well-being.

Contact and Collaboration

We are open to collaboration and further research. If you are interested in working with us or have any questions, feel free to contact the team via the contact page.