4.9.9 TrainOGRLayersClassifier

Train a SVM classifier based on labeled geometries and a list of features to consider.

Detailed description

This application trains a SVM classifier based on labeled geometries and a list of features to consider for classification.

Parameters

This section describes in details the parameters available for this application. Table 4.152, page 795 presents a summary of these parameters and the parameters keys to be used in command-line and programming languages. Application key is TrainOGRLayersClassifier.





Parameter key

Parameter type

Parameter description




inshp

Input vector data

Name of the input shapefile

instats

Input File name

XML file containing mean and variance of each feature.

outsvm

Output File name

Output model filename.

feat

List

List of features to consider for classification.

cfield

String

Field containing the class id for supervision

inxml

XML input parameters file

Load otb application from xml file

outxml

XML output parameters file

Save otb application to xml file











Table 4.152: Parameters table for TrainOGRLayersClassifier.

Example

To run this example in command-line, use the following:

otbcli_TrainOGRLayersClassifier -inshp vectorData.shp -instats meanVar.xml -outsvm svmModel.svm -feat perimeter -cfield predicted

To run this example from Python, use the following code snippet:

#!/usr/bin/python 
 
# Import the otb applications package 
import otbApplication 
 
# The following line creates an instance of the TrainOGRLayersClassifier application 
TrainOGRLayersClassifier = otbApplication.Registry.CreateApplication("TrainOGRLayersClassifier") 
 
# The following lines set all the application parameters: 
TrainOGRLayersClassifier.SetParameterString("inshp", "vectorData.shp") 
 
TrainOGRLayersClassifier.SetParameterString("instats", "meanVar.xml") 
 
TrainOGRLayersClassifier.SetParameterString("outsvm", "svmModel.svm") 
 
# The following line execute the application 
TrainOGRLayersClassifier.ExecuteAndWriteOutput()

Limitations

Experimental. For now only shapefiles are supported. Tuning of SVM classifier is not available.

Authors

This application has been written by David Youssefi during internship at CNES.

See also

These additional ressources can be useful for further information: