jet-c-21/PyCV2GPU

Build python opencv with gpu enabled

PyCV2GPU

Dockerfile Ref from @Borda

Build python opencv with gpu enabled

For ubuntu-18.04 OpenCV-4.1.1 cuda-10.1 cudnn-8

all cuda build

docker image build \
  --build-arg HOST_XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -t cv2-gpu:v6 -f cuda10.1-cudnn8-devel-ubuntu18.04/Dockerfile .

Failed with darknet

docker image build \
  --build-arg HOST_XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -t nd-cv2-gpu:v1 -f darknet_failed_all_gpu_acc_on/Dockerfile .

or cd in it

docker image build -t cv2-gpu:py3.6 .

sudo docker image build -t cv2-gpu:py3.6 .

Container create

• create container

docker run \
  -it \
  --gpus all \
  --net=host \
  -v $(pwd):$(pwd) \
  -v /tmp/.X11-unix/:/tmp/.X11-unix \
  -v /tmp/.docker.xauth:/tmp/.docker.xauth \
  -v $(XDG_RUNTIME_DIR):$(XDG_RUNTIME_DIR) \
  -e XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -e XAUTHORITY=/tmp/.docker.xauth \
  -e DISPLAY=$DISPLAY \
  -w $(pwd) \
  --privileged \
  --name cv2gpu_ctnr1 cv2-gpu:py3.6

• run the container in detach

docker start cv2gpu_ctnr1

• enter in container

docker exec -it cv2gpu_ctnr1 bash -c "cd $(pwd) && bash"

For ubuntu-18.04 OpenCV-4.1.1 cuda-10.1 cudnn-8 (CUDA Capability up to 8.6)

docker image build \
  --build-arg HOST_XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -t cv2-gpu:cv2cc8.6 -f cv2cc8.6-cuda11.7.1-cudnn8-devel-ubuntu18.04/Dockerfile .

For ubuntu-20.04 OpenCV-4.7.0 cuda-11.7.1 cudnn-8 (CUDA Capability up to 8.6)

docker image build \
  --build-arg HOST_XDG_RUNTIME_DIR=$XDG_RUNTIME_DIR \
  -t cv2-gpu:ub20cv2cc8.6 -f cv2cc8.6-cuda11.7.1-cudnn8-devel-ubuntu20.04/Dockerfile .

check

docker run -it --rm --gpus all cv2-gpu:ub20cv2cc8.6 /bin/bash