Add min_distance parameter to peak_local_max call #2008
Conversation
otherwise multiple local maxima are returned even within the distance if their values are identical. This lead to lots of over-segmentation. In particular, multiple perfectly contigous nuclei were split in half. Related spacetx#2006
|
@nickeener @ctcisar do you have any recommendations for the CI fails above? |
|
@berl the failure appears to be in the |
adjust assumption on number of detected cells
|
I found the problem in the iss test. Indeed, the assumption on the number of detected cells had to be updated from 6 to 4. I think this makes sense. The previous pipeline output was this: The new pipeline output is this: So the number of detected cells did change from 6 to 4. |
the docker build was failing, because botocore needs urllib3 to be in a certain version range. Pinning urllib3 to that range fixes the issue.
|
The docker build could not install all dependencies via pip because of this error: Pinning |
|
this is awesome! all that work put in by many people making end-to-end tests of these methods did exactly what it was supposed to. I'll review and merge |
Before this change, multiple local maxima were returned even within the min_distance if their values were identical. This lead to over-segmentation in some cases.
See #2006