sup's profile - activity

I want to draw for example a rectangle bended backwards and project it to a 2D Mat to see effects of the perspective. If no, how can I create 3D Mats with simple shapes ? I have never used graphic engines, what is the simplest approach ?

thanks that makes sense

Can I pass a pointer to Cv::Mat to a C program that will cast it to cvMat ? This is because I want to use a platform(renderscript) that only accepts C code

I don't understand, if clusters it the name of the structure type, shouldn't clusters.covs not make sense ? And also what is newCovs ?

I don't understand, if clusters it the name of the structure type, shouldn't clusters.covs not make sense ? And also what is newCovs ?

OpenMP is simplier since I only need to add one line, what is the advantage of parallel_for ?

I want to access a pixel, which is in uchar format, in double format, for some floating point operations. I am forced to use at<vec3b> since .at<vec3d> will access wrong elements. Is there a way to quickly convert it to vec3d or should I do it manually ?

Also, I thought about using at<mat> but it crashes without informations. Is this behavior normal, i.e Mat not supported ?

Will this for example without any modification ? I am of course developing in native code. Is there a trick compared to the desktop version ?

I am working with OpenCV Android, I need to use a Java OCR library that works on bitmaps. Please, how can I convert my RGBA Mat (which by the way comes from Mat org.opencv.android.CameraBridgeViewBase.CvCameraViewFrame.rgba()) to a Bitmap ? Thanks

EDIT : I think I found the answer in the Utils class. But how to properly initialize the type of the Bitmap object so that it will be compatible with my Mat ?

Will the frame be queued, so if 1 frame is processed per second, after 25 second, we would have processed only one second of the real world ? I want it not to be queued but skipped, so to have a "real-time" feel, even my program can only process 2 or3 frames.

Detail : I am studying the OpenCV Mixed Processing sample.

Thank you very much, I missed that.

The documentation gives no information about their difference.

Hello, I have to write a program to track pedestrians, I thought about :

Using HOG to detect pedestrians in a given frame Associate detected pedestrians from two consecutive frames by euclidean distance. Kalman filter to smooth the trajectory for when hog doesn't work.

Is there an advice you could give me or is this a right approach ?

Thanks