CK <[email protected]> wrote: > Anyway, I used two different algorithms : SVM and KNN. > For SVM I had this message: "(WARNING) Confidence map requested but the > classifier doesn't support it!" > In your mail, you wrote me that I need a model with probability estimates, so > is the issue coming from here? Or is it due to the fact that I'm in 5.6? >
For SVM, you have to set this option on the training step (TrainImagesClassifier): -classifier.libsvm.prob > For KNN, I have an other kind of problem: > Application.logger (FATAL) The following error occurred during application > execution : > /home/mrashad/dashboard/otb/src/Modules/IO/ImageIO/include/otbImageFileWriter.txx:443: > Could not create IO object for file uint16 > Tried to create one of the following: > RADImageIO > BSQImageIO > LUMImageIO > TileMapImageIO > GDALImageIO > MWImageIO > ONERAImageIO > MSTARImageIO > You probably failed to set a file suffix, or set the suffix to an unsupported > type. > Did you correctly set the name of the output image? You need to indicate an appropriate extension so that the application knows which format to use. That is, you should name the output image "myImage.tif" instead of "myImage". Jordi > I've checked my script but it seems good to me... > So, do you know how I can resolve my problems ? > > Thanks in advance, > CK > > Le mardi 14 février 2017 17:22:37 UTC+1, Jordi Inglada a écrit : > > Hi, > > The -io.confmap option does not exist in the TrainImagesClassifier > application. There is an -io.confmatout option to compute the confusion > matrix. > > However, the ImageClassifier application has the -confmap option which > generates what you want: a raster image with the confidence estimated for > each pixel in the classification. The confidence is estimated in a different > way depending on the > classification algorithm. From the documentation: > > " - LibSVM : difference between the two highest probabilities (needs a model > with probability estimates, so that classes probabilities can be computed for > each sample)\n" > " - OpenCV\n" > " * Boost : sum of votes\n" > " * DecisionTree : (not supported)\n" > " * GradientBoostedTree : (not supported)\n" > " * KNearestNeighbors : number of neighbors with the same label\n" > " * NeuralNetwork : difference between the two highest responses\n" > " * NormalBayes : (not supported)\n" > " * RandomForest : Confidence (proportion of votes for the majority class). > Margin (normalized difference of the votes of the 2 majority classes) is not > available for now.\n" > > Jordi > > CK <[email protected]> wrote: > > > > Hi everyone ! > > > > I'm new on OTB (5.6 & 5.8) and I have some questions about the possibility > to get the probability of belonging of a pixel to a class (after or during a > classification). > > > > I tried to use the -io.confmap "option" of the > otbcli_TrainImagesClassifier tools but I only get a matrix on my samples. > > In fact, I would like to obtain a raster on my entire aoi with a value > (between 0 and 1) for each pixel. > > > > I also saw in a previous post on the group the possibility to call the > EvaluateProbabilities. > > I'm not very familiar with this so do you know if there is any other way > to have my probability raster for my classification (using a pre-existing > tools may be) ? > > > > Thanks in advance for your help ! > > C. > > > > -- > > -- -- -- 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.
