niklases

Results 3 comments of niklases

This Dockerfile using ARG CUDA=11.8.0, jax==0.4.14, jaxlib==0.4.14+cuda11.cudnn86, and adding line `\ && conda install -c nvidia cuda-nvcc` works for me (on a single Nvidia RTX 4080, Ubuntu-22.04; Python 3.10 Conda...

You can try to replace the line `code.corners = Mat(4, 4, DataType::type);` with `code.corners = Mat(4, 4, CV_32FC2);` while CV_32FC2 should be the desired cv::Mat type.