CTrainNodeCvRF: serialization problem

[ereator]

The CTrainNodeCvRF::save() / CTrainNodeCvRF::load() seems to not working.
When loading the training data (saved before with CTrainNodeCvRF::save() function) with the CTrainNodeCvRF::load() function, the random forest appears to be empty and the first call of the CTrainNodeCvRF::getNodePotentials() function causes an exception.

However, it is possible to successfully load the training data, which was saved by the DGM v.1.5.0.

In order to reproduce, please insert into the Demo Train the code:
nodeTrainer->save("D:\\");
nodeTrainer->reset();
nodeTrainer->load("D:\\");

after the line
nodeTrainer->train();

[ereator]

After a short investigaton, it is appeared that this bug is related to OpenCV library, which is responsible for serialization. The DGM v.1.5.1 and DGM v.1.5.2 uses OpenCV v.3.2.0, while the DGM v.1.5.0 used OpenCV v.3.1.0. The data files, serialized with the OpenCV v.3.1.0 are still possible to read, using the latest version. The only difference with the data files, serialized with the OpenCV v.3.2.0 is in the file header:

File saved with OpenCV v.3.2.0:

%YAML:1.0
---
opencv_ml_dtree:
   is_classifier: 1
   var_all: 4
   ...

File saved with OpenCV v.3.1.0:

%YAML:1.0
opencv_ml_dtree:
   format: 3
   is_classifier: 1
   var_all: 4
   ...

• This bug is reported to OpenCV: Bug in cv::Algorithm::save / load opencv/opencv#9125
• This bug may be also relevant for the CTrainNodeCvGMM

Quick Fix

Add the string
format: 3
to the saved data file in order to read it with CTrainNodeCvRF::load()

[ereator]

This bug is absent in the newer version of OpenCV v.3.3.0