boil's profile - activity

I'm trying to compile Opencv on my tegra K1 board and don't want to use opencv4tegra. The Ubuntu that is on my tegra has 2 libavcodec libraries installed: /usr/lib/arm-linux-gnueabihf/libavcodec.so.54 /usr/lib/arm-linux-gnueabihf/neon/vfp/libavcodec.so.54

Each time I compile Opencv, it links against the library in the neon folder which I guess contains some optimizations. I would like to link against the other non optimized library. How can I do that?

The reason for all this mess is that I have to decode a mjpeg movie on my intel desktop and on the tegra k1. The problem is that the decoded frames differ between intel and tegra and wanted to test if the avcodec library is doing this.

Thanks

Hi,

I am supposed to port an app to GPU. The app makes extensive use of OpenCV (opencv4tegra). I noticed that HoughLinesP doesn't have the same prototype on CPU and GPU. For example the CPU prototype is this:

CV_EXPORTS_W void HoughLinesP( InputArray image, OutputArray lines,
                           double rho, double theta, int threshold,
                           double minLineLength=0, double maxLineGap=0 );

The GPU implementation has this prototype:

CV_EXPORTS void HoughLinesP(const GpuMat& image, GpuMat& lines, HoughLinesBuf& buf, float rho, float theta, int minLineLength, int maxLineGap, int maxLines = 4096);

The threshold parameter is missing from GPU implementation and because of this I get different results for the same image. Is there an explanation why CPU and GPU functions are not the same? Also why is there no support for streams for HoughLinesP on GPU ?