obtain obj & landrmark with openface

[Inferencer]

I think this is awesome, any chance i can use openface landmark results as the detection is better

but my main question is in the collab the .obj possible to get is as fbx with animation to export to blender is that possible? I could do ply and use an addon but that would mean editing each model each frame

can I also ask if there is a way to remove the background footage in video results?
and if there is a training method for the 3d model so I can fit to face more accurately?