ApplyMixedHaarClassifier & OpenCV cascade

ApplyMixedHaarClassifier & OpenCV cascade

Hello,

I tested the FaceDetection sample of the IPP and it works quiet good.
Now I wrote a converter of the opencv haar feature cascade and an extendet detection code for using the tilted features of opencv.
After testing a lot of parameters, ic got my code running and it detect some faces.
The problem is, that it detect not the right faces. I noticed that the detected faces are on regions in the images, where no edges and a very high illumination is present.
I thought I used all the parameter combinations that are used in the opencv code, but it seems to be not the same, because the cascade I use gives with the performance tool of the opencv code good results.

Is there somebody who tried to use the opencv cascade and can give me some hints what the problem good be.

Thx
Peter

3 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.
Community Admin's picture

Hello, Face detection classifiers have numerical stability problems in case of areas with low contrast. The following can be suggested:

1) use histogram equalization (cvEqualizeHist()in OpenCV) to increase the overal image contrast.

2) playwith CV_DO_CANNY_PRUNING option, and the internal parameters inside cvHaarDetectObjects for earlier rejection of low-contrast areas. AFAIK, IPP sample does not use Canny pruning yet, so you may take the code from cvHaarDetectObjects and adopt it for the IPP-based solution.

I wanted to detect face from a web camera. I found an example of face detection in samples folder.But they have provided a wrapper class called as CIppiImage . I am retriving frame using open cv,so I am not able to figure out that How I will be able to convert Mat Image or IplImage to the type of CIppiImage.

Login to leave a comment.