MarioNETte: Few-shot Face Reenactment Preserving Identity of Unseen Targets(AAAI20)

MarioNETte Architecture

MarioNETte: Few-shot Face Reenactment Preserving Identity of Unseen Targets(AAAI20)_第1张图片
Fig.2展示了MarioNETte的框架图

给定driver image x \mathbf{x} x,一组target images { y i } i = 1 ⋯ K \left \{ \mathbf{y}^i \right \}_{i=1\cdots K} {yi}i=1K,整个framework输出一幅Reenacted image
注意:driver x \mathbf{x} x是一帧图像,target有多幅图像,所以称为few-shot face reenactment

生成器包含以下components:

  • The preprocessor P P P利用3D landmark detector检测人脸关键点,然后将它们render成为landmark image, r x = P ( x ) , r y i = P ( y i ) \mathbf{r}_x=P(\mathbf{x}), \mathbf{r}_y^i=P(\mathbf{y}^i) rx=P(x),ryi=P(yi)

你可能感兴趣的:(读书笔记)