Refine one camera against a fixed existing 3D landmark population. Supplied intrinsics and noisy triangulated points remain fixed, so this is a conditional estimate before joint refinement. Soft-L1 weights use a complete two-coordinate residual norm, with analytic SO3/additive-t derivatives.
Refine one camera against a fixed existing 3D landmark population. Supplied intrinsics and noisy triangulated points remain fixed, so this is a conditional estimate before joint refinement. Soft-L1 weights use a complete two-coordinate residual norm, with analytic SO3/additive-t derivatives.