ZED SDK AND OPENCV INTEGRATION FOR DEPTHSENSING

I have 2 questions

1)We aim to perform real time obstacle detection, using depth sensing which we will normalize and then perform blob detection on, so will a 4gb jetson nano be good enough for this with zed 2i?
2)While I am able to perform and visualize depth sensing and normalization and conversion to opencv form while using a cpu, while using a gpu i get
depth_map_ocv_gpu.download(cpumat) , I get cpumat is an invalid argument in function download , even though I followed the same as in this guide zed-opencv/cpp/src/main.cpp at master · stereolabs/zed-opencv
any idea what could be happening? cpumat is of type cv::Mat and depth_map_ocv_gpu is of type cv::cuda::GpuMat then why do i get this error only while using gpu?

  • Conversion function between sl::Mat and cv::Mat
    */
    cv::cuda::GpuMat slMat2cvMatGPU(Mat& input) {
    // Since cv::Mat data requires a uchar
    pointer, we get the uchar1 pointer from sl::Mat (getPtr())
    // cv::Mat and sl::Mat will share a single memory structure
    return cv::cuda::GpuMat(input.getHeight(), input.getWidth(), getOCVtype(input.getDataType()), input.getPtrsl::uchar1(MEM::GPU), input.getStepBytes(sl::MEM::GPU));
    }

Hi @haadinho10
Welcome to the StereoLabs community

Do you mean the first “old” Jetson Nano, or the new Jetson Orin Nano?
The old version is low power, it allows you to perform obstable detection at low resolution and low frame rate.
Furthermore, it does not support the latest versions of the ZED SDK.

What version of OpenCV and ZED SDK are you using?