At present the rim angle has to be set manually. It might be desirable to have a mode in which an optimal rim angle was derived during the optimisation procedure.
The "titration" method (i.e. trying out reconstructions with a range of rim angles) sometimes gives error surfaces that are bumpy; certainly not convex, so one issue might be avoiding local minima.
Optimising the rim angle at the same time as the point locations would probably require a change in the representation of the coordinates to avoid correlations between the rim angle and the phi coordinate of all mesh points. Another issue would be to consider how to store and annotate retinae that have automatically-generated rim angles.
At present the rim angle has to be set manually. It might be desirable to have a mode in which an optimal rim angle was derived during the optimisation procedure.
The "titration" method (i.e. trying out reconstructions with a range of rim angles) sometimes gives error surfaces that are bumpy; certainly not convex, so one issue might be avoiding local minima.
Optimising the rim angle at the same time as the point locations would probably require a change in the representation of the coordinates to avoid correlations between the rim angle and the phi coordinate of all mesh points. Another issue would be to consider how to store and annotate retinae that have automatically-generated rim angles.