Human detection using svm and hog and counting
6 views (last 30 days)
Hi. I want to find human in a given image and then estimate the number of them.. i need to do this with svm and hog and i dont want to use computer vision . i know how to get hog feature and also how to train svm for simple object detection but i dont know how to use svm to find some people in a image and then locate them with a rectangle?
Dima Lisin on 25 Jan 2016
If you must do this from scratch, then you are going to have to implement the sliding window search. You slide a rectangle of a fixed size over the image, and each time classify it using your HOG-SVM classifier to see if there is a person there.
Or you use vision.PeopleDetector in the Computer Vision System Toolbox, which already includes a pre-trained HOG-SVM model, and will do sliding windows for you.
More Answers (1)
Andrei Cosmin on 1 Feb 2017
Hello there! I want to do same project and I don't know how to start and if the following diagram for the implementation it is ok: - obtain the hog feature; - doing calassification with a SVM code. I have read a lot of papers about the implementation of HOG and SVM and I don't exactly how to apply on my situation. I want to thank you for all response.