MicroGlam: Microscopic Skin Image Dataset with Cosmetics

TL;DR

MicroGlam tackles the challenge of realistic cosmetic rendering by creating a microscopic skin-patch dataset captured with a Light-Stage-inspired device, enabling isolation of cosmetic effects from skin tone and lighting. The authors build a $8\mathrm{mm}\times 8\mathrm{mm}$ patch dataset across $9$ subjects and three cosmetic products (plus no-makeup), with over $600$ images per patch under diverse lighting. They demonstrate a rendering pipeline based on image-to-image translation that renders cosmetics onto unseen patches and compare it to a baseline, achieving greater realism and preservation of skin microgeometry. This dataset enables precise study of skin-cosmetic interactions at micro scales and supports development of more accurate cosmetic-rendering tools for practical applications.

Abstract

In this paper, we present a cosmetic-specific skin image dataset. It consists of skin images from $45$ patches ($5$ skin patches each from $9$ participants) of size $8mm^*8mm$ under three cosmetic products (i.e., foundation, blusher, and highlighter). We designed a novel capturing device inspired by Light Stage. Using the device, we captured over $600$ images of each skin patch under diverse lighting conditions in $30$ seconds. We repeated the process for the same skin patch under three cosmetic products. Finally, we demonstrate the viability of the dataset with an image-to-image translation-based pipeline for cosmetic rendering and compared our data-driven approach to an existing cosmetic rendering method.

Figures (4)

• Figure 1: (a) Side view of our capturing device. (b)(c) Bottom view of the device with all and half of the LEDs selectively turned on, respectively. Each of the $16$ LEDs surrounding the camera center can be switched on/off individually.
• Figure 2: (a) Cosmetic products captured: foundation, blusher, and highlighter. (b) Raw image of a skin patch captured using our device of approximately $8mm^*8mm$. (c) Cropped and aligned version of (b). (d, e, f) The same skin patch but with blusher, foundation, and highlighter applied onto it respectively. We recommend enlarging the figures for a clearer viewing experience. Note how all makeup interacts with the skin microgeometry and affects more than just the diffuse color of the skin (e.g., foundation fills in the wrinkles of the skin and highlighter adds significant specular reflection onto the surface of skin).
• Figure 3: Cosmetic rendering results generated by our image translation-based approach (\ref{['sec:app']}) and the baseline method (kim2018practical). We used the no-makeup images captured at different lighting conditions as input to generate all the results. For each cosmetic product, we provide a reference by presenting the captured skin patch with the exact same cosmetic product applied onto it.
• Figure 4: Following Kim et al., we applied (a) foundation onto a white surface and (b) blusher and (c) highlighter onto a black surface. We use them as input to the baseline method. (zoomed-in images of different cosmetic products)