Crash on depth reconstruction from imported cameras and fixed calibration

[KEgg]

crash on depth reconstruction.

Was running a test importing an existing camera set-up and reconstructing to see if scale can be maintained between image sets / chunks, containing 8 photos per set, without having to align chunks later in the workflow.

First clue to it crashing was its estimation to depth recon time being 20mins, when it was originally doing this process, and its geometry build in less than 7 mins. After almost getting to 100%, it freezes for a while then drops out.

Also noticed that the points generated / matched by the camera alignment was not as dense as if I had done the alignment straight off not using a previous camera set up.



settings used ::

Obj type  : Arbitrary
Target Q  : High
Geo Type : Smooth
Face Count : 50,000
Filter thresh : 0.5
Hole thresh  : 0.1

ubuntu 11.04

Photoscan build 979

[Diego]

Hello,

I also found the same problem in the Windows version, when working with imported camera processing times are vastly greater.  Processing the same data is four or five times slower. I hope that can be solved in the next update.

Thanks for your support.

Regards,

Diego

[Alexey Pasumansky]

Hello Mike,

Thank you for reporting. We'll try to find out what is the problem.

Also noticed that the points generated / matched by the camera alignment was not as dense as if I had done the alignment straight off not using a previous camera set up.

This problem will be fixed in the next update, thanks again.

[KEgg]

Hi Alexey,

Just to say, have tested this function thoroughly on the latest build 985 and it's working brilliantly.

great fix

best regards

Mike