Kao researcher explores machine learning for makeup evaluation with potential for personalisation

By Amanda Lim

- Last updated on GMT

Kao researcher develops machine learning model to objectively analyse and evaluate makeup texture on the skin. [Getty Images]
Kao researcher develops machine learning model to objectively analyse and evaluate makeup texture on the skin. [Getty Images]

Related tags Personalisation Makeup machine learning

Kao researcher develops machine learning model to objectively analyse and evaluate makeup texture on the skin with potential applications in the personalised cosmetics space.

The research utilised deep neural networks (DNNs), a class of machine learning algorithms​ that aims to mimic the information processing of the brain.

It was able to accurately analyse and classify various skin attributes, such as age range, the use of base makeup, and makeup formulation type.

Furthermore, it exhibited the ability to assess makeup conditions and quantify “makeup feel”.

The capabilities of the model make it a valuable tool to help develop makeup that are tailored to individual preferences and characteristics.
The research concluded that further studies need to be conducted to explore its potential in the personalised beauty space.

“This approach has potential applications in visual science research and cosmetics development. Further studies can explore the analysis of different skin conditions and the development of personalised cosmetics.”

More accuracy in evaluations

Conventionally, makeup evaluations were conducted by human experts. However, using DNNs could offer more accurate and objective assessments.

“There have been many reports that various tasks that conventionally had to be judged by humans could be answered with greater accuracy than by humans… We hypothesised that deep learning technology could be used to obtain a makeup finish evaluation technology that can evaluate subtle textures as well as human visual evaluation.”

According to the study, DNNs have previously been applied to skin evaluation. Most notably, it has been used in medicine in the detection of skin cancer.

The study used skin patches – small sections of facial skin images – to train the DNN model.

“The advantages of using skin patches include retaining fine texture, eliminating false correlations from non-skin features, and enabling visualisation of the inferred results for the entire face.”

The DNN model was trained to classify skin attributes and predict the visual assessments of experts.

When applied, the trained DNNs were able to classify various skin attributes, such as such as age range and presence of base makeup. It was also able to identify the difference between liquid and powder formulations.

As a result, the trained DNN could distinguish between bare skin and skin with makeup.

It could also quantify the perceived makeup feel and visualise its distribution across the face.

In the study, makeup feel refers sensation of makeup on the skin.

The model was able to identify the makeup feel with the application sunscreen and it continued to increase with the subsequent application of primer as well as foundation.

Additional evaluation tasks showed the model’s ability to provide insights into various skin texture aspects.

“The trained DNNs on regression task showed high prediction accuracy for the experts’ visual assessment. Application of DNN to the evaluation of actual makeup conditions clearly showed appropriate evaluation results in line with the appearance of the makeup finish.”



Source: Skin Research & Technology

Skin patch based makeup finish assessment technique by deep neural network

Author: Ken Nishino


Related news

Show more

Follow us


View more


Beauty 4.0 Podcast