Best face landmark with jetsons?

2 years ago I had tried “dlib+opencv+python” in various jetsons. The project was not successful, as the fps was not fast.

Now I am so happy that there is fpenet with landmarks and it can give face landmark coordinates in python, without using opencv and dlib and native nvidia solutions instead…

Yet as a real amateur, I am struggling to find the best way to reach the best setup that gives me landmark coordinates in python.

I hace a new build jetson nx. What should be the terminal command records (for a newbie:( like me) to build the setup to reach the point to catch the face landmark recognizing with v4l2 usb camera. ( docker? deepstream? Tao? Tensor rt? Jetpack? Ngc? Fpenet deployable version? Jupyter?all?)

It would be a big help if I would knew the steps 🙏

Best regards