I have a query regarding the scaling of the detector. I have implemented the algorithm in MATLAB and able to detect the faces of different sizes by resizing the image itself. However as per my study of the topic, we have to scale the detector itself instead of scaling the image. A classifier has properties such x, y , width, length , posMean (Mean of thresholds computed of positive examples which are classified correctly), similarly posMin, posMax , range j etc. It is clear that in scaling , we have to multiply the x,y,width and length of the classifier with the scaling factor, but i have till now come to know that what to do with the rest of the properties which are responsible for setting the range or the criteria for a particular classifier to Pass/Fail a subwindow as face or non-face.
Can anyone help in this regard? Thanks.
Submit an article or tip