First draft of structural similarity class #88
Conversation
...ava/org/hortonmachine/hmachine/modules/statistics/kerneldensity/OmsStructuralSimilarity.java
Outdated
Show resolved
Hide resolved
| @Execute | ||
| public void process() throws Exception { | ||
|
|
||
| GridCoverage2D meanMap1 = windowMean(inMap1); |
There was a problem hiding this comment.
I see you tend to use gridcoverages and convert them to HMRasters once you need them.
It is best that you create an HMRaster at the very begin of the module for all the maps you need and then pass these to the various window* methods.
Then at the end of the modeul you build the output GridCoverages from the HMRaster as you need.
There was a problem hiding this comment.
Hi @moovida thanks for the comments! I'll implement your suggestions later in the morning
There was a problem hiding this comment.
I implemented the changes. Now all the window functions take HMRasters as arguments and also return HMRasters instead of a GridCoverage2D. I don't know if you see an issue with that. My thought process is that as soon a GridCoverage that will be certainly used is produced I convert it to HMRaster
There was a problem hiding this comment.
Yes @diegomvd , apart of the return from the methods, which I commented, it looks fine).
Your thought was ok, but the conversion between from GC to HMRaster has a cost. So it is best to use HMRaster (which has all the methods you need) and then convert ony at the end.
| throw new ModelsRuntimeException(errorMessage, this); | ||
| } | ||
|
|
||
| return HMRaster.fromGridCoverage(outputRaster.buildCoverage()); |
There was a problem hiding this comment.
Here just return outputRaster, it is a HMRaster already.
@diegomvd , creating this PR for the purpose of having a discussion board.