So,I was reading the paper visualizing object detection features by antonio toralba
visualizing hog features
In this the author mentions there are situations in which the descriptor results in false positives.
Unfortunately the author doesn’t give much information about why this happens ,apart from mentioning that it has something to do with the design
So i wanted to know what are some of the problems or scenarios in which histogram of oriented gradients might fail…
Also,there are other descriptors like SIFT,SURF,Daisy etc.So how are these better than HoG both in terms of robustness and efficency.
Also,if anyone is aware of any survey that compares these descriptors that would be helpful.