Recognition and classification of skin condition

The beauty industry offers tens of thousands of cosmetics and treatments. An individual approach to consumers’ beauty needs is key to maintaining their interest.

Our client - one of the world’s largest manufacturers of beauty products – was looking for a tool that automates recognition and classification of the clients’ skin to measure its attributes, such as evenness or hydration.

deepsense.ai created a system that analyzes photos of skin to automate the process of suggesting cosmetics. The system recognizes and classifies the skin tone and evenness of the client’s face. The assessment is preceded by very careful data preparation including face detection, normalization and controlled adaptation through various image augmentations. A deep neural network then processes the photos to provide the skin grading. The solution produced robust outputs despite the varying photo quality and lighting conditions, making it suitable for working with studio-quality photos and hand-made selfies.

Our solution is 91% accurate*, significantly outperforming humans. Further improvements to be made to the system may include the ability to recognize other customers’ skin attributes such as hydration, shine, pores, acne scarring and wrinkles.
*measured by mean average error