Thanks Guillaume - Would it be possible to get the raster image and the training data? Is it something from the OTB-Data directory? If so, which files exactly?
If I have the same data, I can then try to run the same classification in R, and then compare the model outputs of both OTB/OpenCV and R. On Thursday, 11 December 2014 12:56:21 UTC-5, Guillaume Pasero wrote: > > If you want to have a look to the models generated by OTB, I have run > the application *TrainImagesClassifier* on a simple case (5 classes, > input image is a Pleiades XS extract). > I have attached the different output files obtained after training for > different classifiers coming from OpenCV. > > I am not familiar with this type of format, but I guess it can be > translated. > > Regards, > Guillaume > > Le 11/12/2014 17:13, dcd a écrit : > > Merci Guiallaume, > > It looks like the OpenCV saves models in XML or YAML format. In R, I > believe there's a way to translate models built in R to PMML, which should > make it possible to save into XML or YAML. So, in terms of the file format, > it seems possible to communicate between the two applications. > > However, the point I think you're alluding to, is how to figure out the > actual differences in the model formats written by OpenCV and R. > > Perhaps I can take a look and compare the output from R (model -> PMML -> > XML/YAML) with one generated by OTB? Do you know where OTB save it's model > files? > > I'm sure this will be much more complicated, but it would be such a > TREMENDOUS feature: the power of R with OTB would be pretty much the best > thing ever for remote sensing image processing. > > > > > > > On Thursday, 11 December 2014 03:58:20 UTC-5, Guillaume Pasero wrote: >> >> Hi, >> >> I think this feature isn't so far away. At the moment, in OTB >> classification framework, once we have trained a model, it is saved in a >> text file. The classifier called next opens the file, load the >> corresponding model, and use it. >> Thanks to OpenCV, we support different classifier model (such as RF). The >> export format of the model is defined by OpenCV. >> >> It means that if you manage to store your model using the right format, >> it can be imported and used in OTB. >> >> You can check here : >> >> http://docs.opencv.org/modules/ml/doc/random_trees.html?highlight=rtrees#cvrtrees >> http://docs.opencv.org/modules/ml/doc/statistical_models.html >> >> Regards, >> Guillaume >> >> Le 11/12/2014 04:55, dcd a écrit : >> >> >> I'm back and using OTB again and I was wondering if there's a way to use >> models developed in R, with their counterpart classifier in OTB? In other >> words, if I've trained a RF model in R, is there some way to export that >> model to OTB's RF classifier? >> >> I'm guessing it's not possible (yet), but it would be a killer feature as >> it would provide the best of two worlds: OTB's raw processing power and >> speed with R's sophisticated model building abilities. >> >> I'm not sure how the best way would be to go about doing this, but the >> following links might be of interest: >> >> Export RF model to SQL >> https://gist.github.com/shanebutler/96f0e78a02c84cdcf558 >> >> Random Forest: PMML Export and ADAPA Support >> >> https://support.zementis.com/entries/21214261-Random-Forest-PMML-Export-and-ADAPA-Support >> >> Cheers, >> >> Dennis >> >> -- >> -- >> 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. >> >> >> >> -- >> <http://www.c-s.fr> *Guillaume PASERO* >> Ingénieur d'études et développement >> *Business Unit E-SPACE & Geo Information* >> <https://thor.si.c-s.fr/blogs/cs-blogs-business/>* - Département >> APPLICATIONS* >> >> *CS Systèmes d'Information* >> Parc de la Grande Plaine - 5, Rue Brindejonc des Moulinais - BP 15872 >> 31506 Toulouse Cedex 05 - FRANCE >> +33 561 17 64 21 - [email protected] >> > -- > -- > 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] > <javascript:> > To unsubscribe from this group, send email to > [email protected] <javascript:> > 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] <javascript:>. > For more options, visit https://groups.google.com/d/optout. > > > > -- > <http://www.c-s.fr> *Guillaume PASERO* > Ingénieur d'études et développement > *Business Unit E-SPACE & Geo Information* > <https://thor.si.c-s.fr/blogs/cs-blogs-business/>* - Département > APPLICATIONS* > > *CS Systèmes d'Information* > Parc de la Grande Plaine - 5, Rue Brindejonc des Moulinais - BP 15872 > 31506 Toulouse Cedex 05 - FRANCE > +33 561 17 64 21 - [email protected] <javascript:> > -- -- 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.
