Bug in GpuMat convertTo with CV_16F #27373
Description
System Information
Version : OpenCV 4.12b
OS : Ubuntu 22.04
Compiler : g++ (Ubuntu 12.3.0-1ubuntu1~22.04) 12.3.0
CUDA : 12.9
GPU : NVIDIA RTX 5090
Detailed description
The method convertTo is buggy on cv::cuda::GpuMat on CV_16FC1.
Steps to reproduce
cv::Mat test32(1, 2088960, CV_32FC1);
cv::Mat test16;
test32.convertTo(test16, CV_16FC1);
// works
cv::cuda::GpuMat test32(1088, 1920, CV_32FC1);
cv::cuda::GpuMat test16;
test32.convertTo(test16, CV_16FC1);
// works
cv::cuda::GpuMat test32(1, 2088960, CV_32FC1);
cv::cuda::GpuMat test16;
test32.convertTo(test16, CV_16FC1);
// error: (-217:Gpu API call) an illegal memory access was encountered in function 'call'
Issue submission checklist
- I report the issue, it's not a question
- I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
- I updated to the latest OpenCV version and the issue is still there
- There is reproducer code and related data files (videos, images, onnx, etc)