A transformer based on source keypoints and target keypoints
As some TPS_STN released by other contributers, they are transformer by source grid points and target grid points.
The basic idea is from
[1](blog talk with this kind of image transfomer)[https://blog.csdn.net/VictoriaW/article/details/70161180] [2]the paper talk about this [http://user.engineering.uiowa.edu/~aip/papers/bookstein-89.pdf]
A project may use the same idea (https://github.com/menorashid/animal_human_kp)
from this
to
