Is it possible to make HoG scale invariant?
I was wondering if its possible to make hog features scale invariant by using a normalization method?
I was wondering if its possible to make hog features scale invariant by using a normalization method?
Asked: 2016-03-22 09:18:00 -0600
Seen: 1,039 times
Last updated: Mar 22 '16
How to match 2 HOG for object detection?
Question regarding feeding extracted HoG features into CvSVM's train
How to estimate the noiselevel of an image?
How to find a match between 2 shifted hue histograms?
Many CvSVM vectors into one... how? [closed]
Single Image Depth Map / Sharpness Map
How to match two images and find out mistakes
Training of SVM classifier in OpenCV using HOG, SIFT and ORB features
Remove buttons from their background so I can re-use button image
What do you mean by scale invariant? This is a histogram, so it is by definition.
I did not knew HoG was scale invariant. By scale invariant, I mean that for example I trained an object Object_1 of size A when i am testing an object of size 2*A. Will it classify it as Object_1?
For geometrical scale invariance you have to resize your region to a specific size otherwise you can get a different number of features from HoG. Use cv::resize with a fixed size for this.
I shold resize the ROI, that is I put a bounding box over it, crop it and then resize it to a specific size.