Hi David,

let's try to debug the error with OTB 6.7. Indeed to use random forests
classifier the correct way to set it is with '-classifier rf' option. And
Random forests has not been removed in the last version and the way it is
set from the TrainImagesClassifier did not change neither. In this case the
random forests algorithm from opencv is used internally.

Can you run otbcli_TrainImagesClassifier without arguments to display the
application documentation and tell me what classifier are listed as
available in the documentation?

It can also  be related to a configuration issue with the superbuild. Can
you check that you compile and install OpenCV with the superbuild?

Can you post your CMakeCache.txt file? and also the output of the OTB
configuration (you can find it in
superbuild_build_dir/OTB/src/OTB-stamp/OTB-configure-*.log?

Regarding OTB unit test, OTB is test on a nightly basis on Win/Linux/Mac.
The log from the tests are available publicly:

https://dash.orfeo-toolbox.org/index.php?project=OTB

I don't see currently issues with the TrainImagesClassifier but I perhaps
miss something (I've tested on my platform that the test related to random
forests currently passed).

Let's check your configuration and see



Le mer. 12 sept. 2018 à 14:37, <davidmichea.unis...@gmail.com> a écrit :

> Hi,
>
> I have a problem with otbcli_TrainImagesClassifier.
>
> I have 2 version of OTB installed :
>
>    - one is 5.10.1 delivered in binary package for ScientificLinux 7.2
>    - the other is the 6.7.0 build from source (Superbuild) yesterday
>
> I launch this command to use a random forest classifier on my dataset :
> numactl --cpunodebind=1 --membind=1 otbcli_TrainImagesClassifier -io.il ./
> 9_bands_2020200_ATTRIBUTES.tif -io.vd ./INPUT/9_bands_2020200_SAMPLES.shp
> -sample.mv 10000 -sample.mt 10000 -sample.bm 0 -sample.vtr 0.5 -sample.vfn
> class -classifier rf -rand 1 -io.out ./rfModel.txt -io.confmatout ./
> rfConfusionMatrix.csv
> (forget about numactl: not relevant here)
>
>
>    - when I use the 5.10.1 version, I got this error message :
>
> 2018 Sep 12 13:36:54  :  Application.logger  (INFO) Sampling rates...
> 2018 Sep 12 13:36:54  :  Application.logger  (FATAL)
> 2018 Sep 12 13:36:54  :  Application.logger  (FATAL) Missing arguments in
> strategy.constant.nb to process sampling rates
> 2018 Sep 12 13:36:54  :  Application.logger  (FATAL) The following error
> occurred during application execution :
> /home/otbval/Dashboard/src/OTB/Modules/Applications/AppClassification/app/otbMultiIm
> ageSamplingRate.cxx:250:
> itk::ERROR: Fatal error caught. Aborting..
>
> And when I try to pass the -strategy.constant.nb parameter, I get that :
> ERROR: option -strategy.constant.nb does not exist in the application.
>
>
>    - when I use the 6.7.0 version, I got this message :
>
> 2018-09-12 13:49:36 (FATAL): itk::ERROR: ChoiceParameter(0x7e9980):
> Invalid parameter value 'rf'
> given for parameter '-classifier'
>
> Did you remove the random forest algorithm from the classifier ?
>
>
> I implement on a large computing facilities a code that have been tested
> and implemented with the 5.2 version of OTB. But for example, the
> otbcli_Segmentation function was not working properly with the 5.10.1
> version, that is why I have compiled the last version from sources.
>
> But I wonder how can I now which function works or not for different
> versions as several working functions seems to be broken in further
> versions ...
>
> Have you something like a dashboard online where I could check
> functionnalities and unit testes for each function of OTB for each version ?
>
> Best regards,
> David
>
> --
> --
> Check the OTB FAQ at
> http://www.orfeo-toolbox.org/FAQ.html
>
> You received this message because you are subscribed to the Google
> Groups "otb-users" group.
> To post to this group, send email to otb-users@googlegroups.com
> To unsubscribe from this group, send email to
> otb-users+unsubscr...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/otb-users?hl=en
> ---
> You received this message because you are subscribed to the Google Groups
> "otb-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to otb-users+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>


-- 
Manuel Grizonnet

-- 
-- 
Check the OTB FAQ at
http://www.orfeo-toolbox.org/FAQ.html

You received this message because you are subscribed to the Google
Groups "otb-users" group.
To post to this group, send email to otb-users@googlegroups.com
To unsubscribe from this group, send email to
otb-users+unsubscr...@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/otb-users?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"otb-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to otb-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to