Home Number of steps in the training of each stage of the haar cascade is very high (OpenCV)
 I trained a cascade classifier with OpenCV, but for each stage it didn't reach the suitable False Alarm Rate. A part from the training process is shown below:  ===== TRAINING 7-stage ===== Training until now has taken 0 days 0 hours 8 minutes 41 seconds.  About 2000 of samples are the real world samples and others are created by opencv_createsample tools. I don't know exactly what is the problem. I guess the augmented images is very hard for discrimination, is it true?