In general the Remote Module system works fine on Windows, but I don't know about boussaffawalid/FeatureSelection in particular.
Note that we are currently working on making a binary package for remote module, but unfortunately boussaffawalid/FeatureSelection will not be included :( Victor Poughon ________________________________ De : [email protected] [[email protected]] de la part de Jake Shermeyer [[email protected]] Envoyé : vendredi 27 janvier 2017 16:34 À : otb-users Cc : [email protected] Objet : Re: [otb-users] Band Importance in Multi-band Raster Image Classification Thanks guys! Will check out both of these options. Maybe a silly question as I haven't investigated the remote modules closely, but will they work in a windows environment or is Linux recommended? Cheers! On Friday, January 27, 2017 at 3:54:02 AM UTC-5, Jordi Inglada wrote: Hi Jake, You may try to approach your problem more systematically using the feature selection module which is part of OTB's remote modules https://github.com/boussaffawalid/FeatureSelection This will however need that you compile your own OTB in order to activate this module. If you don't feel like doing that, but you are comfortable with some scripting (bash, python), you can try to loop over different band combinations and use a cross-validation approach to select the best combinations of input features. I hope this helps, Jordi Jake Shermeyer <[email protected]<UrlBlockedError.aspx>> wrote: > > Hi all, > > I'm presently using the Gradient Boosting Tree classifier as part of the > TrainImagesClassifier tool. I have a vector training dataset and a multi-band > raster that serves as my input. I've been successfully outputting models and > cal/val confusion matrices. I then use these matrices as verification data > to optimize my classifier, adjust parameters, and tweak the bands I use as > inputs. > > As part of this process, I've been adding and removing bands included in my > input raster and I would like to see which bands are contributing or are > weighted the most in the classification. > > For example the classifier makes decisions and groups pixels into specific > classes based on each band x% of the time: > Band 1- 10% > Band 2- 25% > Band 3- 50% > Band 4- 15% > > Having this information will allow me to remove less valuable bands and > ensure I'm using the strongest set of data that will be the most effective. > > Is this a possibility? Or is there a comparable work-around if not? > > Thanks > > -- -- -- 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 [email protected] To unsubscribe from this group, send email to [email protected] 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 [email protected]<mailto:[email protected]>. For more options, visit https://groups.google.com/d/optout. -- -- 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 [email protected] To unsubscribe from this group, send email to [email protected] 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 [email protected]. For more options, visit https://groups.google.com/d/optout.
