Hi everyone. I'm trying to create samples from hand images to recognize hand. When I want to create samples from an image there is no problem and api is creating (sample count defined with -num parameter and image file with -imf) *.vec file but when I want to create samples from images collection (with positive images file and name info parameter -info and again with -num) api's result is (for example; I've 500 positive image and I'm giving 1000 for -num and my image info file at /opt/hand/pos.txt);
/opt/hand/pos.txt(501) : parse errorDone. Created 500 samples
when I change the -num parameter to 500 it's just write to terminal Done. Then I train the classifier for -numStages 10 but train process ended at 4th stage (but the output is not an error here the link 3rd answer is saying). Is this normal (I think not because if -num X and -img just one X samples will be created but if -info has 10 images mustn't it be 10X)? I'm using OpenCV2.4.8 on Ubuntu 12.04. At the last I've tried to get capture from webcam and tried to recognize the hand but program was few things without hand ( but it detect my face :) I think it probably about the positive and negative images count. Thanks.