Lecture 27: Inverse Graphics/Research In Angjoo's Group (62)
sethzhao506
How do we predict the right arm if that's no appearing in the image? Would it be deviated from the reality when reconstructing into 3D scene for this?
philippe-eecs
I would assume you would infer, I think this is where a prior would be super important. A 2d image would not give you enough information to reconstruct a 3D object, we'd need hefty priors on the human shape to know that there should be a right arm present behind the person. Or we'd need multiple pictures with different views.
How do we predict the right arm if that's no appearing in the image? Would it be deviated from the reality when reconstructing into 3D scene for this?
I would assume you would infer, I think this is where a prior would be super important. A 2d image would not give you enough information to reconstruct a 3D object, we'd need hefty priors on the human shape to know that there should be a right arm present behind the person. Or we'd need multiple pictures with different views.