Consider a single foreground object being observed by a set of converging cameras positioned on a circle looking inward $($e$.$g$.$ the object is at the center of said circle$).$ The angle of separation between neighboring cameras is one degree $($i$.$e$.$ we have $360$ total cameras$)$ If you were only allowed to use a subset of $36$ cameras to estimate the observed $3$D geometry, how would you choose the most favorable $($in terms of reconstruction accuracy$)$ camera subset under each of the following two reconstruction scenarios:

running shape$-$from$-$silhouettes from the binary foreground segmentation of each input imagery.

running multi$-$view stereo based on photoconsistency.

Assume the cameras are already calibrated and their pose perfectly known. Justify your answers and make a sketch of your proposed selection schemes