Is there currently a bug with nppiLabelMarkers_8u32u_C1R in cuda 10.1 windows?
When I process an 8u image, it tells me the number of markers is 6000 but when I look at the result, there are 6000 different numbers but the majority of the numbers are the same as the 8u image and only a small amount of the pixels are the 6000 different numbers. The 8u image has only values of 0 and 255, but the output image that has values of 255 are not touching each other.
There was a recent issue reported here:
but that should be fixed in CUDA 10.1
Probably can’t say much without a complete test case that shows the issue. If you’re not using 10.1 U1 (10.1.168) you may wish to try that. If you have a complete test case, you’re welcome to file a bug at developer.nvidia.com. The instructions are linked to a sticky post at the top of the CUDA programming sub-forum.
Note, bug was submitted https://developer.nvidia.com/nvidia_bug/2662787
Hi, Have this problem resolved? I have met the similar problem. Ｗhen I used nppiLabelMarkers_8u32u_C1R, some different connected areas are labeled with the same number.
You’re welcome to file a bug if you wish, as the previous poster did.
For general comments, those having difficulty with the current NPP label markers functions are advised to wait for the next CUDA release (after CUDA 10.1) and use the UF label marker functions there.
https://docs.nvidia.com/cuda/npp/group__image__filter__label__markers.html
I don’t have any information on the next CUDA release, schedule or otherwise.
I won’t be able to provide further instructions or details here.
This is the current response that I got (UF was also broken in 10.1)
“This issue will be fixed in the new UF functions in upcoming NPP release. We have no plans to fix this in the old non-UF functions. Also the new UF functions are MUCH faster than the old functions.”
Hi,
i get error ( not connected labels is the same )
using nppiLabelMarkers_8u32u_C1R
here’s the log: npp_label.log - Google Drive
Checked cuda versions (10.0,10.1,10.2)
Driver versions ( 4.18, 4.40)
Graphics card: GeForce 940MX
( bug submitted )
https://developer.nvidia.com/nvidia_bug/2865865