System Information

OpenCV version: 4.6..0
Operating System / Platform: Ubuntu 24.04
Compiler: GCC 13.2.0

Detailed description

A value of the nu parameter to SVM ml algorithm (type NU_SVC) that is inconsistent with the distribution of labels of the training cases triggers a silent failure in the train() method. The problem is detected, but no error or warning printed--it just fails. Suggested change would be to replace the "// TODO" line with something like :

CV_Error( cv::Error::StsBadArg, "training cases incompatible with nu parameter--try a lower value");

The same issue exists in the current source at the time of this submission.

Steps to reproduce

The dog that did not bark is on line 1462 of module/ml/svm.cpp :

                        if( nu*(ci + cj)*0.5 > std::min( ci, cj ) )
                            // TODO: add some diagnostic 
                            return false;
                    }

https://github.com/opencv/opencv/blob/11c69bb1717e2e0f37b2c25c4741aa4b18538788/modules/ml/src/svm.cpp#L1460

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
• I updated to the latest OpenCV version and the issue is still there
• There is reproducer code and related data files (videos, images, onnx, etc)