My goal is to detect a certain mood and it's absence based on my facial features. So the idea was to first detect my face and have only the facial expression of the mood that I want, use those pictures as the positive pictures and then on the flip side have negative pictures that show me not having this mood. Then to use both of them to train a classifier. The current command line (64bit) looks like this opencv_traincascade.exe -data YesTC -vec Mood.vec -bg NonMood.txt -numstages 30 -minHitRate 0.999 -numPos 120 -numNeg 120 -maxFalseAlarmRate 0.5 -w 20 -h 20 -precalcValBufSize 1024 -precalcIdxBufSize 1024.
The good thing is, it does not crash, which happened a lot of times before.
the bad thing is, after some stages it aborts, saying
Train dataset for temp stage can not be filled. Branch training terminated.
I tried using relative paths as suggested at http://answers.opencv.org/question/10872/cascade-training-error-opencv-244-train-dataset-for-temp-stage-can-not-filled-branch-training-terminated-cascade-classifier-cant-be-trained-check-the/, I tried absolute paths, I tried saving it as unix file, always the same result.
How to fix this?