How is the face mask application implemented?

Question

Android

I want to try to make an application with different effects for face selfies. For example, makeup, masks, glasses, etc. I can not understand how such applications are implemented? I googled, there is TensorFlow, there is ARCore (Augment reality). How such effects are implemented in Instagram, Snapchat and VK. With neural networks? Does anyone know what is the correct way to use it? Is it difficult to implement this?

Answer 1

The computer vision library finds the position of the face. A mask is applied over the anchor points of the tour. Most do not take into account the angles of rotation.
Everything is in open source, including a trained model for face search. Clones are baked every month. The task for a beginner is interesting, but the prospects for earning more than 1000 bucks on this are illusory