|
|
|
|
|
|
by fromthestart
2542 days ago
|
|
|
You train the neural net to encode landmarks which are unique to individual faces. Such a net, when trained, can extract the unique facial features from a single image and embed them in a vector to be compared to features extracted from other images. This is similar to how single images of faces can be animated with GANs now. https://arxiv.org/abs/1905.08233v1 |
|
|