liquidmetal's profile - activity

Matching feature points directly won't work, as you've already discovered. There are two issues that you want to solve.

1) Use a narrower field of view. You can reproject pixels from the spherical map onto an image plane with a narrower fov. This would produce regular looking images on which you can use the standard feature matching techniques, computing the fundamental matrix, etc. You're looking for the inverse spherical projection equations for this.

2) Register the two panoramas. It is likely that the two panoramas are relatively shifted. Is the object at the left-most edge of the first panorama the same as in the second one?

Hopefully this helps!

I'm trying to recreate something like the Flash demo on this page. Computing homographies between image pairs is straightforward.

I want to use the computed homographies to create a display like this using Viz. Any suggestions on how to go about doing this?

Thanks!

I'm trying to write my own cuda kernel but I can't find a way to pass an array of cv::GpuMat to the cuda kernel.

Here is what I have until now:

__global__ void somename(cv::cuda::PtrStepSz<float> *array_of_images) {
    ...
    }
...
int main(...) {
    cv::cuda::GpuMat *mats = new cv::cuda::GpuMat[180]();
    // Initialize eachof the mats
    somename<<<...>>>(mats);           // Doesn't work
}

The compile error that the above snippet produces is:

filane.cu(89): error: argument of type "cv::cuda::GpuMat *" is incompatible with parameter of type "cv::cuda::PtrStepSz<float> *"

What's the correct way to pass an array of images to a cuda kernel?

I want calculate run cv::sum on the GPU - so I looked to cv::cuda::sum. However, I'm running into issues. This github gist has all the details. I've launched

https://gist.github.com/liquidmetal/9...

Any help on how to fix this would be appreciated. I suspect I may be doing something wrong.

find_package requires this additional file code.opencv.org/attachments/196/FindOpenCV.cmake. This requires that I set the variable OpenCV_DIR. Is there a way to locate OpenCV without setting this variable?

I'm trying to update the cmakefile for another project that depends on OpenCV. What's a good way to identify the correct OpenCV version across multiple platforms? I just need to figure out if OpenCV 3 is available on the current machine or not. Any hints or ideas on how to do this would be useful!

Here's a summary of what I'm trying to accomplish:

• run locate libopencv_imgcodecs | head -n 1 | wc -l
• If this command returns 0, assume OpenCV is installed (I know this is potentially a bad assumption)
• If the command returns a non-zero number, set OpenCV version to 3.
• Set the variable OPENCV_VERSION based on this.

I don't feel too good about running the shell command (locate, etc).. is there a better way to do this?

This file seems to be missing from Github now. Was it removed from the project?

https://github.com/Itseez/opencv/tree...

Let's say I'm running kmeans on a dataset with a termination criteria on 100 iterations and epsilon of 0.01. Once the function returns, how do I find out which termination criteria was used to terminate the algorithm?

Did anyone find a way to do this yet?