Hey everyone, I am wondering about the decision to have cv::ocl::convertFromImage copy the image data into a new OpenCL buffer owned by a cv::UMat whereas cv::ocl::convertFromBuffer simply attaches the cl_mem buffer to a cv::UMat. Is there some reason that OpenCV cannot do the same for an OpenCL image2d? Unfortunately you can’t use convertFromBuffer on an image2d because it fails type and size assertions when attempted. Is this merely a case of not being needed yet, or is there some deeper issue that would render this impossible without a copy? Thanks.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| OpenCV and iGPU shared memory | 2 | 1793 | November 11, 2022 | |
| How to initialize cv::UMat by using data pointer? | 4 | 692 | November 8, 2023 | |
| Why does calling ORB_Impl::detectAndCompute block when using cv::Umat | 0 | 598 | October 28, 2021 | |
| Usage of UMat and Mat | 6 | 4884 | April 24, 2023 | |
| Why does _InputArray::convertTo(OutputArray dst, int rtype, ..) or cv::convertTo(..) not exist? | 2 | 796 | November 29, 2021 |