I am looking for advice on the best way to track a face for a research project. Model targets work for object recognition of the whole face when scanned with a 3d scanner and trained on that scan. But we would like tracking to work when the mouth and nose are obscured.
I have also tried using image targets with 3 QR codes placed on the forehead and cheeks which seems to not work as well as the model target.
I have looked into using MultipleTargets but they seems to require a cuboid shape with images taking up the entire surface area of the cuboid's faces. This might work if the image to be recognized could take up only a small portion of the cubiod.
I am wondering if anyone with experience can recommend a method of enhancing object recognition in the scenario I have described. One thing I am now considering is enhancing the 3d scan of the face with stickers that might aid in model target recognition over a smaller area?