Dusting of images #107
-
I'm using this package to find spatiotemporally contiguous heatwaves in remotely-sensed land surface temperature data (MODIS-LST). I have created a binary 3D (latitude, longitude, time) array, where heawave days are 1, and non-heatwave days are 0. Due to the nature of the dataset, it is quite noisy, so the individual heatwaves aren't always connected. The connected_components method finds way too many small heatwaves (which are actually part of a bigger heatwave). However, when I try dusting the array to get rid of the small clusters, I get a "TypeError: No matching signature found" if I choose a threshold other than 1 (which does not change the array at all). Do you have any idea why the dusting might not work for me with a threshold other than 1? Any suggestions how to solve this problem? |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 8 replies
-
Hi! Thanks for using cc3d! I'm very happy you're finding interesting uses for the software. I suspect that you may be using an old buggy version of dust. Can you let me know what OS, Python version, and cc3d version you're using? I tried running dust on a binary noise array with types uint8 and bool and didn't trigger this error. Note that Python3.6 and below are EOL and do not download current versions of cc3d from PyPI. Upgrade your Python version in that case. |
Beta Was this translation helpful? Give feedback.
-
As noted in the previous discussion, dust currently only accepts unsigned ints in 3.10.4. I'll be releasing 3.10.5 which expands the acceptable types to signed ints. |
Beta Was this translation helpful? Give feedback.
As noted in the previous discussion, dust currently only accepts unsigned ints in 3.10.4. I'll be releasing 3.10.5 which expands the acceptable types to signed ints.