- See also
- python.workflows.clfCreateTrainingPatches
Aim of module
Create training patches for deep learning neural networks.
General description
More details to come...
- See also
- Script documentation
Parameter description
-patchgeometry string definition for patchgeometry.
Type : String
Remark : mandatory
Description: This defines the geometry within which points are searched for.Possible examples of calls are "circle(r=50)" or "window(x=50 y=50)".
-classAttribute attribute which contains the class labels
Type : String
Remark : mandatory, default: ['Classification']
Description: This parameter determines the attribute that the classification is based onand that points are selected from. Usually this will be "Classification".
-classIds Points of which class Ids should be selected for training.
Type : String
Remark : optional, default: ['2=2', '5=5', '6=6', '18=18', '30=30']
Description: The class Ids should be provided in the following format: "classId(int)"="class_description(str)".If not specified, all classIds of the input file(s) are takeninto consideration.
-useAttributes determines which attributes to export.
Type : String
Remark : optional, default: ['Id', 'Amplitude', '_EchoWidth']
Description: attributes can be specified that are desired to be exported in the training patches. If not specified, only the minimum set of attributes are exported. Namely 'x', 'y', 'z' and 'Classification'. Setting this parameter to 'all' leads tothe exporting of all available attributes of the odm.
-excludeBorder whether or not border points are included in the random selection
Type : String
Remark : optional, default: auto
Description: This determines, how far a point has to be from the border in order for it tobe considered for selection. Per Default it is fitted to the size of the buffer,but can be set to 0 or to a desired value.
-filter modify the input using a (tree of) filter(s)
Type : String
Remark : optional
Description: Reduce and/or transform the input data set: import last echoes only, transform coordinates, modify attributes, etc.
-patchesPerClass determines the number of patches generated per class.
Type : Integer
Remark : mandatory, default: ['2']
Description: per class specified in the parameter 'classIds' the number of patches specified viathis parameter is generated. If 5 classes are given with 10 patchesperClass, 50 patcheswill be generated in total.
-exportFormat specify output format of the point clouds.
Type : String
Remark : optional, default: .las
Description: point clouds are exported in the format provided by the user. Currently,las, shp and odm exports are supported
Common Package/Script Options
Settings concerning the general options for (package) scripts.
-infile input text file, directory or strip file
Type : Path
Remark : mandatory
Description: As input either the name of an ASCII file (extension .txt) containing the names of the data files (one file per line) or an expression using wildcards to specify the respective data sets are accepted.
-outfile output directory or settings file (.txt)
Type : Path
Remark : optional
Description: Specifies the directory where the final results are stored. If not specified, the results are created in the current working directory.
-tempdir temporary directory
Type : Path
Remark : optional, default: current directory
Description: Specifies the directory where all intermediate results are stored. If not specified, a TEMP subdirectory is created in the current working directory. Additional subdirectories (one per involved module) are created.
-cfg Path to configuration file
Type : Path
Remark : optional, default: $OPALS_ROOT/cfg/clfCreateTrainingPatches.cfg
Description: The name of a configuration file containing all relevant calculation parameters is expected. If not specified, the default cfg files as provided by the OPALS distribution are used.
-projectDir project directory
Type : Path
Remark : optional, default: current directory
Description: The default project directory is the current working directory, but can be easily changed by this parameter. All path parameters, if not specified as absolute paths, are interpreted relative to the project directory.
-skipIfExists skip processing if result already exists
Type : Boolean
Remark : optional, default: True
Description: Skip processing if result already exists. In order to re-run current script it is useful to repeat the processing only if the respective output does not already exist. This allows for incremental processing of large projects.
| possible input | evaluates to |
| 1, true, yes, Boolean(True), True | Boolean(True) |
| 0, false, no, Boolean(False), False | Boolean(False) |
-distribute determine how many local processes should be used
Type : String
Remark : optional
Description: according to the number chosen as the value, independent processes are being started to parallelise (parts of) the program and therefore enhance the runtime.
Logging Options
Settings concerning the verbosity level of logging.
-fileLogLevel Log level in the logfile
Type : LogLevel
Remark : optional, default: info
-screenLogLevel Log level on screen
Type : LogLevel
Remark : optional, default: info
-logger Logger
Type : Logger
Remark : optional
Description: Logger is usually provided by the opals framework.The user may provide their own logger object, but it has to function in the same way as the opals Logger.
Examples
To be written.
1.8.17