Tweet
In datasets, it is often the case that some regions are much better "covered" with camera views than others.
It is currently already possible to adjust the number of cameras by which a point needs to be viewed to increase accuracy ("point confidence") which does seem to make a difference up to five cameras.
When set to the highest value, the "low coverage areas" are often too sparse. When lowering the number of cameras, the "well covered regions" are sampled unnecessarily densely.
It might be of interest to think of a way of adjust the number of cameras required for point definition dynamically, based upon a sort of "heat map" of camera visibility.
Don't know. It is just a thought.
It is currently already possible to adjust the number of cameras by which a point needs to be viewed to increase accuracy ("point confidence") which does seem to make a difference up to five cameras.
When set to the highest value, the "low coverage areas" are often too sparse. When lowering the number of cameras, the "well covered regions" are sampled unnecessarily densely.
It might be of interest to think of a way of adjust the number of cameras required for point definition dynamically, based upon a sort of "heat map" of camera visibility.
Don't know. It is just a thought.