Video pre-processing step

[subharya83]

I see the pre-generated crops of faces from individual frames are currently resized to N x N.

Will aligning the faces in the same video clip using the facial landmarks that are output during the landmark detection help make a better landmark generator?