Image: Hair segmentation.
We’ve developed a series of computer vision solutions to study the quality of Henkel’s new hair care products (QA process). These products – hair dyes, hair conditioners, and hair mascaras – undergo extensive quality controls before they reach the market. The quality control process consists in analyzing the color obtained after dyeing, the resistance to washing, and homogeneity, shine, flexibility, and movement of the hair among other key points.
Thanks to our automated tool, the QA process can be performed accurately and objectively in a fraction of the time required to do it manually.
Image: User interface tool with controls and color measurements.
Before, these processes were performed manually by direct observation. Different products were applied to people’s hair -or standardized hair for experiments- and they were evaluated by a group of hair treatment experts. In addition to the time required by these experts to assess the suitability of each new formula, Henkel was facing the problem of individual and non-objective assessment.
Thanks to our computer vision algorithms, we study color, homogeneity, behavior, flexibility, gloss and other parameters instantaneously and objectively.
A specific algorithm has been developed for each parameter. Convolutional Neural Networks (CNN) for object segmentation is used to delimit the hair strands of interest. In addition, automatic color calibration and white balancing algorithms are used to obtain the most accurate range of colors. Regression machine learning models are used to estimate color homogeneity, as well as various distance metrics in color spaces such as CIE L*a*b*, to resemble human color perception as closely as possible. Computer vision techniques based on geometric estimations have also been employed to detect key points within the hair strands treated with these innovative products.
(1) Source: URL [www.henkel.com/company]
Henkel logo © 2022 Henkel AG & Co. KGaA. All rights reserved.