WIP: Add class probability output for RF classifiers
Summary
Add an output to the ImageClassifier application providing an image with the probability of each class for each pixel.
Rationale
The image has as many channels as the number of classes and each channel stores the probability of a class in the order of the labels. Only the implementation for SharkRF has been done.
Implementation Details
Classes and files
Many files and classes are touched, since the approach is to add a tag which indicates if a particular ML model knows how to output the probabilities. It is the same approach used for the confidence map. Therefore, all ML models have now this tag.
The ClassificationFilter has a new output containing the probability map.
Applications
The ImageClassifier application has a new parameter "probamap" to activate the output. The number of classes has to be given also, since not all models store the number of classes.
Additional notes
This MR still needs work, but I am submitting it in order to have feedback.
Copyright
The copyright owner is COPYRIGHT OWNER (OR OWNER'S AGENT) and has signed the ORFEO ToolBox Contributor License Agreement.
Check before merging:
- All discussions are resolved
- At least 2
👍 votes from core developers, no👎 vote. - The feature branch is (reasonably) up-to-date with the base branch
- Dashboard is green
- Copyright owner has signed the ORFEO ToolBox Contributor License Agreement