Didn't want to file a bug report about this but thought I'd share it anyway.
Short version: RC has trouble aligning difficult datasets and can crash when (too many?) Control points are present. For the solution, read the rest... :-)
I am currently working on a very difficult dataset (c400 images@24mpx), a building which is surrounded by trees and bushes. Without any CPs, I get about 10 components with 3 to 150 images, so not great. To help the alignment along, I started falling into the control-point-trap, meaning I started somewhere and all of a sudden there were 100+ :O
At the beginning, everything went great and with each group of CP more and more images got aligned. The alignment took from 3 to 12 minutes, depending on whether there were already some components or not, so quite fast thanks to RC.
At some point, the alignment would not finish anymore. One time I left it running and went out of the house. When I came back, it was running since 3,5 hours (!!!) and nothing was happening anymore. After some trying, I figured out that some CPs in especially difficult areas, where RC would sometimes get things confused, seem to be causing this behaviour. After I added another batch of CPs, now RC would crash instead of just hang. It only happens, when there is not prior component present in the project. As long as something is there to give RC some idea of what's happening, everything is fine. So this is the solution, to deactivate all CPs but a few essential ones, run a first alignment and then, with some components present, enable the remaining CPs. Interestingly this all happens around 30 to 40 % of the last processing step.
Please sign in to leave a comment.