Vladislav Vinogradov's profile - activity

Yes, it looks like MOG was removed from 3.0.

You need to manually convert array of GpuMat to array of PtrStepSz:

cv::cuda::GpuMat *mats = new cv::cuda::GpuMat[180];
// Initialize eachof the mats

cv::cuda::PtrStepSz<float> h_ptrs = new cv::cuda::PtrStepSz<float>[180];
for (int i = 0; i < 180; ++i)
    h_ptrs[i] = mats[i];

Then you need to copy h_ptrs to GPU memory:

cv::cuda::PtrStepSz<float> *d_ptrs = NULL;
cudaMalloc(&d_ptrs, 180 * sizeof(cv::cuda::PtrStepSz<float>));

cudaMemcpy(d_ptrs, h_ptrs, 180 * sizeof(cv::cuda::PtrStepSz<float>), cudaMemcpyHostToDevice);

Then you can pass d_ptrs to kernel:

 somename<<<...>>>(d_ptrs);

I submitted a fix for the issue PR.

Did you install CUDA Toolkit? Try to set CUDA_TOOLKIT_ROOT_DIR CMake variable.

Use IMREAD_GRAYSCALE and IMREAD_COLOR flags for imread calls:

Mat image = imread("1.jpg", IMREAD_COLOR);
Mat mask  = imread("mask.jpg", IMREAD_GRAYSCALE);

cv::gpu::compare(B, cv::Scalar::all(k), A, cv::CMP_GT);

GpuMat doesn't support expressions like Mask == 0. You have to call plain functions like cv::gpu::bitwise_not:

cv::gpu::bitwise_not(mask, mask);
img.setTo(Scalar::all(0), mask);

You are building master branch (3.0 version) of OpenCV, not 2.4.8.

Please add more information about your issue. Do you see any error message?

You can't access oclMat data from CPU code. The oclMat data is located on GPU memory. You can only upload/download from/to CPU Mat or run custom OpenCL kernel.

Yes, SurfFeatureDetector supports only CV_8U depth. Also it will convert color image into grayscale.

GPU filters don't work in-place. Use different GpuMat objects for input and output parameters:

cv::gpu::GpuMat gpuinput=...;
cv::gpu::GpuMat gpuoutput;
cv::gpu::filter2D(gpuinput,gpuoutput,gpuinput.depth(),ker);

• temporalAreaRadius - the number of images used in the super-resolution.
• iterations - the number of iterations in steepest descent solution for super-resolution minimization problem.

How did you build this example? Did you build it as a part of OpenCV project or you added farneback_optical_flow.cpp to your project? Did you modify this source? opencv2\gpu\device\saturate_cast.hpp header can be used only in CUDA code that compiles with NVCC, but farneback_optical_flow.cpp doesn't include this header.

GPU filters doesn't work in-place.

Currently, super resolution module is disabled for iOS, because it wasn't tested on the system. You can modify CMake script of the super resolution module (opencv/modules/superres/CMakeLists.txt) and removed the first three lines:

if(ANDROID OR IOS)
  ocv_module_disable(superres)
endif()

This will allow you to build superres module for iOS.