Object detection using Opencv Haar Cascades

[Seida]

Hello,
I was wondering if there's someone here to help me, as I'm in some kind of a mess.
I've implemented my own AdaBoost detector, but as the training process was going to take too much of my time, I tried to use opencv's haarcascade_default_face.xml.
I converted it into a txt file, and I read the data, I loaded it into my structures. I began scanning the image with windows 24x24, then I scaled them with 1.25...and nothing, it won't detect any face.
I guess there is something wrong, maybe with the use of the variance. I mean I calculated the variance for the window and I multipled it with the threshold of each feature, compared the feature value (calculated with the integral image) with this threshold, and added left_val, or right_val, according to the case. I compared the resulting sum with the stage threshold.
I must mention that the variance was calculated using the integral of the square image.
I've also rescaled the location of each feature, by simply multipling each rectangle coordinate with the scale.
Thank you