The Potential of Artificial Intelligence for Makeup
Technologies:
About the Client
Our client was a company developing mobile applications using Artificial Intelligence.
Business Context
The client wanted to develop an application that could integrate with video communication platforms such as GoogleMeet, Zoom, etc., and that would broadcast on people’s faces, makeup from the template they specified. In simple words, you upload the image of the desired make-up to the application, and at the output you get an image from two cameras – the original and the modified one.
Thus, the main task was to develop an application that would help people save time for additional preparations before important online events, and at the same time feel natural and comfortable.
Amazinum Data Scientists In Action
Our Data Scientists developed an application that broadcasted the image from the photo to the virtual camera of the final client. For this, we chose the Google MediaPipe model. It made it possible to mark key points on a person’s face, which are responsible for certain features. With the help of the model and triangulation, our specialists achieved a more accurate and reliable superimposition of the photograph on the faces of people in individual parts. At the same time, it was possible to use different photos for different parts of the make-up.
Among the difficulties faced by our specialists was the possibility of a natural and realistic look of make-up on the forehead. After all, there was a big risk that when turning the head, or changing the position of the lips or eyebrows, the makeup could come off and look unrealistic. For this, Amazinum specialists experimented with various models and settled on MediaPipe. It enables triangulation and avoids overlapping. Instead, with the help of points, the model naturally determined the boundaries and features of a person’s facial features for the correct translation of the image from the photo.
