object recognition using HOG features.

asked 2015-06-11 03:05:13 -0500

tomnjerry gravatar image

I am extracting HogFeatures to extract information from images. These numerical values i.e Hog features are used to train SVM for object classification. Is it the correct way to make a classifier for object recognition ? Also, how do I go about having the SVM trained.

edit retag flag offensive close merge delete


Yes it is, there is a very nice example right here! Just do not forget to apply the HOG feature normalisation step before training an SVM.

StevenPuttemans gravatar imageStevenPuttemans ( 2015-06-11 07:51:25 -0500 )edit

@StevenPuttemans based on your experience, do you think it's so important to normalise HOG features?

LorenaGdL gravatar imageLorenaGdL ( 2015-06-11 10:16:51 -0500 )edit

@Lorena GdL I did not perform exhaustively testing on the SVM interface, but when I did, all the SVMs trained with that normalisation for pedestrian detection just outputted rubbish. That is when I found out about the normalisation and since then, my own models have been working. I guess it is necessary for OpenCV, not sure for other implementations. Maybe something in the internal optimization of the SVM that is influenced by this?

StevenPuttemans gravatar imageStevenPuttemans ( 2015-06-12 01:42:54 -0500 )edit