Bareminerals skin care Concept Test

Year: 2014


Client: Bare Escentuals

Location: New York


Role: Design and Development of the experience, in association with Future Colossal

Description

The final concept for this project was to give its users advice on what products among the ones offer ed by Bare Escentuals would better suit their skin tone.

The main technological challenge was to guarantee the effective and continuous skin tone recognition for each person under variable lightning conditions and using hardware of uneven power, since every person would carry the recognition process in their own home. To placate the client’s worries, a concept test was built, trying to recognize more than 30 different skin tones and then determine the best make up for each one of them.


Gallery


Challenges

Generating uniform images independently from the camera used and the ligthning conditions

Due to the use of different kinds of cameras and to the variation in lightning conditions under which the skin color analysis would be made, it was necessary to define strategies to get images with uniform colors, trying to mitigate the effects of the environmental variations. To achieve this, color correction and image processing techniques were applied; these guaranteed acquiring images with constant characteristics, so it would be possible to make homogeneous and consistent recommendations.

Getting efficient facial recognition

Skin tone varies not only among people. The same face may have different tonalities depending on the zone being examined. To be able to make accurate recommendations, it’s necessary to know which portion of the face has the tone to work on. It was then necessary to apply object detection and face recognition techniques to determine the face position, and then extract additional features such as lip, eyes and nose positions.

Determining skin color

Color representation technologies normally used on computational applications (RGB) don’t offer enough fidelity to express the subtleties in tonal changes in a person’s skin. It was then necessary to use a different model (HSV), with which it is possible to represent more variation in skin tone, so the offered recommendations would be more accurate.

Generating consistent recommendations

It was evident from the beginning that the center of the experience would lie in the recommendations offered by the system. It was then necessary to use supervised learning techniques to create and refine the rules that assigned categories to each skin tone, so it would be possible to recommend a kind of make up that would be really appropriate for it.


Used Technologies

Face recognition.  To be able to detect the main features of the faces, and so define key regions to detect the skin color.

Supervised machine learning. To classify skin colors in an effective and consistent way.

Image Processing. To acquire uniform images independently from both the webcam used, and illumination conditions.


Results

Even though the client decided not to move forward with the project, the results obtained through the concept exploration made by the team at Zooloop made clear its technical viability, and became another tool in the technological arsenal of the company.


Copyright 2015 Zooloop | Todos los derechos reservados | Zooloop