Hi I am currently working on training the SVM for HOG using SVM light. Following the methodology of the paper I am able to get decent results but not as good as using the vector being returned by the DefaultPeopleDetector. I am currently trying to reproduce the results from the paper. Does anyone know how exactly the DefaultPeopleDetector was generated so that I can compare that with the Dalaal 2005 paper results as well as my own results. I am mostly interested in whether or not the SVM was trained using a RBF kernel vs a linear kernel and the method of training.
This forum is disabled, please visit https://forum.opencv.org