New Model constructed using VOC takes more time in Latent SVM

asked 2013-10-30 06:22:16 -0500

kwasif gravatar image

updated 2020-10-17 10:31:44 -0500

I have developed a new person detector using pascal VOC 4.0 with my own dataset. I also managed to convert it to xml format as required for opencv latent svm. The problem now is previously the standard xml file provided by opencv extra takes 22 sec for a 1440*1080 resolution image while performing person detection. For the same image the newly generated xml from pascal voc 4.0 takes 77 sec. Both the xmls have same number of features and filters. I do not understand what is the problem here? Any help would be appreciated?

edit retag flag offensive close merge delete

Comments

You got same winstride and block size in both models during training? What features were used?

StevenPuttemans gravatar imageStevenPuttemans ( 2014-04-25 06:41:22 -0500 )edit

Please how did you generated new model?

EricFokou gravatar imageEricFokou ( 2015-05-06 03:35:52 -0500 )edit

http://www.cs.berkeley.edu/~rbg/latent/ --> like reported in the other topic

StevenPuttemans gravatar imageStevenPuttemans ( 2015-05-07 04:58:41 -0500 )edit