Intensity Interpolation, sometimes fast some others crash

[PablerasBCN]

When I interpolaty SF from one cloud to another, I use the simplest, "nearest neighbour" option. Sometimes works suepr fast, some others simply stalls .
CC. Same occurs with color interpolation.

Edit:

For instance. Interpolating from One very dense cloud to a sparse cloud and viceversa allways works and in a fast fashion.

Between 2 dense clouds (20M). it may work or it may stall.

[daniel]

I guess it is indeed highly dependent on the source cloud size and density...

[PablerasBCN]

But I'm not so sure of that.

I'll try to prepare some samples. I just happened to stall it again.

Relatively super fast color transfer on higher density part and stalls on the lower density part. I would spect to go slower but not stall. Also CPU is 100% when this happns.

I'll post samples here when ready

[PablerasBCN]

aaaargg

CC just crashed doint a silly color interpollation of a tinny cloud from a bigger cloud. lost 1h color correction :\\\\\\\\\

I think there has to be some kind of small bug here. Like it does not clear a buffer of some short and keeps growing till overflows or something like that.

[daniel]

If you can provide us with a reproducible example, we could probably fix a potential bug.

[fhaberland]

I am having the exact same problem. I use the color interpolation from another entity to colorize a SLAM pointcloud. I get two clouds from the SLAM processing software. One dense and one colorized. The colorized version only keeps the points that were associated with images and ends up loosing some important goemetry because the scanner (and the camera) were not pointed in that direction. So I use the interpolate function to colorize the points in the dense file that overlap with the colorized file. To avoid crashing, I slice both clouds in smaller areas (about 20.000 points) and then colorize each section. Even then, if the area covered by the dense cloud is larger than the colorized area, the system crashes. CPU goes to 100% and it takes hours to process 20.000 points. It seems that the likelihood of crashing is associated more with the distance between points than with the number of points.

[daniel]

Which version are you using? On which OS?