VideoProcessingFramework

VPF stands for Video Processing Framework. It’s set of C++ libraries and Python bindings which provides full HW acceleration for video processing tasks such as decoding, encoding, transcoding and GPU-accelerated color space and pixel format conversions.

VPF also supports exporting GPU memory objects such as decoded video frames to PyTorch tensors without Host to Device copies. Check the Wiki page on how to build from source.

Docker Instructions (Linux)

1. Install docker, docker-compose and nvidia-docker following the official instructions for your distribution. You can find your distribution as follows.

lsb_release -a
Distributor ID:	Ubuntu
Description:	Ubuntu 20.04.3 LTS
Release:	20.04
Codename:	focal

2. Install dependencies

# the basics
sudo apt-get update
sudo apt-get install git build-essential python3 python3-pip python3-virtualenv
sudo apt-get install apt-transport-https ca-certificates curl gnupg-agent software-properties-common

# python poetry
curl -sSL https://raw.githubusercontent.com/python-poetry/poetry/master/install-poetry.py | python3

3. Set path to Video_Codec_SDK

# Done once
export VIDEO_CODEC_SDK=<path_to_your_Video_Codec_SDK>
export DOCKER_BUILDKIT=1
export COMPOSE_DOCKER_CLI_BUILD=1
cp -a $(VIDEO_CODEC_SDK) Video_Codec_SDK

4. Build & Run image

docker-compose -f docker/docker-compose.yml build vpf
# Get test sample
wget http://www.scikit-video.org/stable/_static/bikes.mp4 -P $HOME/Downloads/
# run image
docker-compose -f docker/docker-compose.yml run -v $HOME/Downloads:/Downloads vpf
# or this way
docker run  -it --gpus=all -e NVIDIA_DRIVER_CAPABILITIES=video,compute,utility -v $HOME/Downloads:/Downloads nvidia/videoprocessingframework:vpf
python Tests.py 0 /Downloads/bikes.mp4 /Downloads/bikes-vpf.mp4

5. Build & Run image with pytorch extension

docker-compose -f docker/docker-compose.yml build --build-arg GEN_PYTORCH_EXT=1 vpf
# Get test sample
wget http://www.scikit-video.org/stable/_static/bikes.mp4 -P $HOME/Downloads/
# run image
docker-compose -f docker/docker-compose.yml run -v $HOME/Downloads:/Downloads vpf
# Run predictions on video
python SampleTorchResnet.py 0 /Downloads/bikes.mp4

6. Build & Run image with OpenGL extension

docker-compose -f docker/docker-compose.yml build --build-arg GEN_OPENGL_EXT=1 vpf
# Get test sample
wget http://www.scikit-video.org/stable/_static/bikes.mp4 -P $HOME/Downloads/
# run image
docker-compose -f docker/docker-compose.yml run -v $HOME/Downloads:/Downloads vpf
# Render video
python SampleOpenGL.py --gpu-id 0 --encoded-file-path /Downloads/bikes.mp4

You can build tensorrt enabled image by replacing vpf with vpf-tensorrt in the above steps and test the following.

python SampleTensorRTResnet.py 0 /Downloads/bikes.mp4

Documentation

cd docs
make html

In case doc building scripts run into isses finding PyNvCodec or PytorchNvCodec modules, add them to PYTHONPATH like shown below:

#assuming your CMAKE_INSTALL_PREFIX is /home/user/Git/VideoProcessingFramework/install
export PYTHONPATH=/home/user/Git/VideoProcessingFramework/install/bin:$PYTHONPATH

Community Support

If you did not find the information you need or if you have further questions or problems, you are very welcome to join the developer community at NVIDIA. We have dedicated categories covering diverse topics related to video processing and codecs.

The forums are also a place where we would be happy to hear about how you made use of VPF in your project.