How could I go about getting access to the exact frame for which the landmarks were calculated?
In several machine learning problems getting those points are crucial for properly formatting the input.
But it doesn't look like there's an easy way to accurately access it.
The obvious choice would be to use a texture provider on the frame update event but there's no guarantee the landmarks will be for that frame.
Instead, one idea I have is to create a new Camera and continuously update its render target with the frame that comes from the update script event and then call facial landmarks. This way I know which frame the landmarks are for.