Ask Your Question

any possible to detect object with cascade training model and HOG feature

asked 2017-01-13 01:20:06 -0500

Link Qian gravatar image

I'm an newbie to OpenCV 3.2.0. recently I train a object detector via opencv_traincascade with HOG feature type. Then I write the code for detecting object.

  HOGDescriptor hog_desc;
  if (hog_desc.load("hog-cascade.xml")) {
     Mat src, src_gray;
    src = imread("test.jpg");
    cvtColor(src, src_gray, COLOR_RGB2GRAY);
    equalizedHist(src_gray, src_gray);
    vector<rect> foundLocations;
    hog_desc.detectMultiScale(src_gray, foundLocations);    

the last line of code raises the error OpenCV Error: Assertion failed( dsize.area() > 0 || (inv_scale_x > 0 && inv_scale_y > 0)) in size, file .../opencv/modules/imgproc/src/imgwarp.cpp, line 3493

my question is HOG detector was removed from opencv 3.2?

edit retag flag offensive close merge delete



it seems, you're confusing HOG cascades and the HOG descriptor.

berak gravatar imageberak ( 2017-01-13 01:30:55 -0500 )edit

HOG feature was remove quietly from cascade classifier without any comment in 3.2.0 release note.

Link Qian gravatar imageLink Qian ( 2017-01-15 19:03:46 -0500 )edit

1 answer

Sort by ยป oldest newest most voted

answered 2017-01-13 01:40:33 -0500

berak gravatar image

updated 2017-01-13 01:41:37 -0500

you're out of luck with hog cascades in opencv3, those were quite silently removed from the CascadeClassifier.

you also cannot use those with the HOGDescriptor, it's a different model (SVM).

so, either train a HOGDescriptor , or an LBP or HAAR Cascade

edit flag offensive delete link more


great answer! thank a lot!

Link Qian gravatar imageLink Qian ( 2017-01-13 21:43:16 -0500 )edit
Login/Signup to Answer

Question Tools

1 follower


Asked: 2017-01-13 01:20:06 -0500

Seen: 307 times

Last updated: Jan 13 '17