Ask Your Question

Revision history [back]

First of all I would suggest using the newer traincascade algorithm for HAAR training which has the newer C++ like interface.

Secondly I wonder why there is no -info parameter in your command, linking to the positives.txt file, which contains the annotated regions for positive training samples in your training data.

Go through this guidelines for haartraining and see if it works out:

Or like I said, try the newer interface.

Also, try to be more concise in your questions next time? You are not giving us any relevant information.