When it comes to next generation retail, it is important to segment client types.
Some clients need recommandations and human contacts while others are fine or prefer with self-service.
The luxury industry is no exception to this trend and our daily digital world make the consumers more autonomous in their purchase experience. Now, the approach must be more subtile in the luxury industry than in plain retail.
In order to split the consumer flow into these two categories, LM3LABS has developed and deployed an interactive kiosk solution for self makeup for a famous cosmetic brand in Japan.
The solution lets the user select her language. Then, she selects a look among the list which instantly appears on her face as a realistic makeup. She can unselect items from the look. Once the selection is made, she can print the list of the items and give it to the sales person.
The Self Makeup solution works with the same quality on ladies with glasses and on all skin tones. Indeed, LM3LABS developed a unique blending algorithm which mimics the real makeup powder. The colored dust creates the same natural result on any type of skin.[/vc_column_text]
The solution extensively used Deep Learning for creating face models. The tracking algorithm was trained on averaged faces of asian women, then weighted by Chanel shop visiting probability based on current trends. This weighting was critical in defining the precision of the tracking as well as the accuracy of the colors on various asian skins.
The model was also trained on tens of hours of YouTube videos to fine-tune motions and random expressions. The below movie illustrates this learning phase at mid-project. The colors used in this movie are not calibrated and the focus is to train the machine learning algorithms on the various possible expressions.
From Deep Learning training to store deployment. A big stretch and a lot of necessary communication and education work.