Problem with opencv_traincascade version OpenCV 2.4.10.0

edit

I added this fix myself to ensure people do not run into the following situation

• Training a first model
• Checking the resulting detector
• Adapting training parameters and retraining a second model
• Noticing that model 2 still uses configuration parameters from model 1

In this case, without cleaning the folder containing the in between models several parameters will be read from the parameter file in that folder. In the situation above it is thus more suitable to first delete that folder or use a new empty folder for the second model training.

The approach of not deleting the folder is ONLY good if you want to continue training by for example adding more stages!