Ask Your Question

What are the literature papers on which traincascade is based?

asked 2015-03-16 06:03:03 -0500

fatih gravatar image


I am working on cascade classifiers and trying to understand the details of the implementations in "apps/traincascade" folder of the source code. In, three papers are mentioned about Haar like features and LBP.

I found "Zhu, Fast Human Detection Using a Cascade of Histograms of Oriented Gradients, 2006" paper seems to be related with HOG implementation. Is this paper correct? Do you suggest any other papers (can also be related with boosting)?

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2015-03-16 07:22:36 -0500

updated 2015-03-16 07:22:49 -0500

Basically you have

  1. First paper described the original approach by Viola and Jones on boosted cascades of weak classifiers to get a strong classifier. However that paper only uses a small set of basic upright features.
  2. Then the second paper by Rainer describes an extended set of features, by rotating the features over 45 degrees and enabling more features from a given window to be used for classification. This resulted in a know accuracy improvement of trained models for several object classes, however not for all possible classes.
  3. Is the paper that describes how the LBP features are implemented in OpenCV.

Aside from that users of the OpenCV community took the liberty of implementing a cascade classifier interface, also known as boosting for the LBP and HOG features. There is no paper exactly describing the approachm but the Histogram of Oriented Gradients is best described in the Dallal and Triggs paper where they combine it with SVM classification instead of boosting.

edit flag offensive delete link more


And let me add that digging deeper into boosting will probably be overkill. Keep in mind that boosting is actually something created in mathematics to solve very complex computationally heavy problems. Better read papers that use the technique in applications interesting to you.

StevenPuttemans gravatar imageStevenPuttemans ( 2015-03-16 07:25:28 -0500 )edit

Thank you very much.

For HOG implementation, I found the following papers with similar approaches: 1. "Porikli F., Integral Histogram: A Fast Way to Extract Histograms in Cartesian Spaces" -> Describes the method to compute histograms (not histograms of gradients) of all possible target regions in an image. 2. "Said, Human Detection Based on Integral Histograms of Oriented Gradients and SVM" -> Extends the method in 1 for computing integral histograms of oriented gradients to be used with SVM. 3. "Zhu, Fast Human Detection Using a Cascade of Histograms of Oriented Gradients" -> Another study, extending 1 for again computing integral histograms of oriented gradients, to be used with AdaBoost.

fatih gravatar imagefatih ( 2015-03-16 10:45:21 -0500 )edit

Currently, I do not have any idea about the similarity of the methods in paper 2 and 3 between each other and to the implementation in opencv. I just wanted to share.

fatih gravatar imagefatih ( 2015-03-16 10:47:36 -0500 )edit

Question Tools



Asked: 2015-03-16 05:56:48 -0500

Seen: 297 times

Last updated: Mar 16 '15