Cuda opencv in docker

Accelerated Computing Intelligent Video Analytics DeepStream SDK
,
1

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) AGX Xavier/1080Ti
• DeepStream Version 5.0
• JetPack Version (valid for Jetson only) 4.4
• NVIDIA GPU Driver Version (valid for GPU only) 440.100

Refer to my question here, I am developing deepstream plugin in a docker container (docker pull nvcr.io/nvidia/deepstream:5.0-20.07-triton) running on my PC using 1080Ti on ubuntu 18.04. Then deploy to my AGX Xavier. I recently installed cuda opencv on my AGX Xavier to enable gstdsexample plugin to run opencv filter in cuda. (see the jetson cuda opencv installation script here)

Question: is there a script that I can install the cuda opencv (or equivalent) in my docker container that run on dGPU? such that I can develop deepstream plug in in my PC instead and once mature, deploy to AGX Xavier. Thank you very much for your help

3

is there a script that I can install the cuda opencv (or equivalent) in my docker container that run on dGPU?

could try below script to build the opencv-3.4.0

 $ mkdir build
 $ cd build
 $ cmake -D CMAKE_BUILD_TYPE=RELEASE   -D WITH_CUDA=on -D ENABLE_FAST_MATH=1 -D CUDA_FAST_MATH=1 -D WITH_CUBLAS=1 -D WITH_NVCUVID=on -D CUDA_GENERATION=Auto   -D WITH_OPENCL=off -D WITH_OPENCL_SVM=off -D WITH_OPENCLAMDFFT=off -D WITH_OPENCLAMDBLAS=off -D BUILD_opencv_cudacodec=OFF ..
 $ make -j8
 $ make install
4

Thanks for the info. And try it. It fails with following error message:

root@06836899da24:/home/cuda_opencv/build# cmake -D CMAKE_BUILD_TYPE=RELEASE   -D WITH_CUDA=on -D ENABLE_FAST_MATH=1 -D CUDA_FAST_MATH=1 -D WITH_CUBLAS=1 -D WITH_NVCUVID=on -D CUDA_GENERATION=Auto   -D WITH_OPENCL=off -D WITH_OPENCL_SVM=off -D WITH_OPENCLAMDFFT=off -D WITH_OPENCLAMDBLAS=off -D BUILD_opencv_cudacodec=OFF ..
CMake Error: The source directory "/home/cuda_opencv" does not appear to contain CMakeLists.txt.
Specify --help for usage, or press the help button on the CMake GUI.
5

what’s your exatc steps to build OpenCV? Please share as detail as you can

6

Thank you for the following up.

Now I am able to install cuda opencv on the docker using exactly the same script.

1 Like
8

Hi ynjiun, could you share your dockerFile? I’va issues to compile opencv with the script. My dockerfile is:

FROM nvcr.io/nvidia/deepstream-l4t:5.0-20.07-samples
COPY install_opencv4.3.0_Jetson.sh .
RUN ./install_opencv4.3.0_Jetson.sh /user/local/src

To RUN the script I have to delete all “sudo” in the file and add unzip the install list.
Everything goes well but CMAKE faild in the process
Thanks a lot! Hernán.

9

did you install cmake?

The following are the steps I install my cuda opencv:

  1. edit install_opencv4.3.0_Jetson.sh to remove all “sudo” and uncomment “make install”
  2. apt install zip unzip
  3. apt install cmake
  4. mkdir build
  5. ./install_opencv4.3.0_Jetson.sh build

I hope it helps.

10

Thanks a lot ynjiun :).
I forget to launch the docker build process with the runtime=nvidia :). Now its complete the process.
Thanks!