Skip to content
forked from uwdb/visualroad

Visual Road: A Video Data Management Benchmark

License

Notifications You must be signed in to change notification settings

ne1114/visualroad

 
 

Repository files navigation

Setup Procedure on Azure NV6 VM

Create Azure Virtual Machine

  • Recommend NV6
  • Make sure to add external storage during the creation process (256GiB recommended)
  • Follow this link after creating the VM to enable the external storage
  • Recommend changing the owner and group of the directory you mounted disk to $USER

Install Pre-requisites

  • Follows the steps 2 ~ 4 in this link

Modify docker configuration and runtime environment

  • Stop to docker using
    sudo service docker stop

  • Modify daemon.json file in /etc/docker/ (If it doesn't exist create one) as following:

{
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    },
    "default-runtime":"nvidia",
    "data-root":"/path_to_mount_directory"
}
  • "data-root" : Pass the path to the mount directory that has 256GiB

  • "default-runtime" : Setting environment variable to use nvidia runtime by default

    • This enables docker's access(?) to GPU
  • Start docker
    sudo service docker start

  • Run docker info and check if the output is similar to below

 Default Runtime: nvidia
 ...
 ...
 Docker Root Dir: /path_to_mount_directory

Follow the instructions in VisualRoad main repo from Step4

  • Install ue4-docker
sudo pip3 install ue4-docker
sudo ue4-docker setup
  • Build ue4 docker image In order to build the ue4 docker image, you need to create EpicGames account and link/connect(?) to github account Once you created the account do following steps:
ue4-docker build --cuda=9.2 --no-minimal --no-full --exclude=debug --exclude=templates 4.22.0
  • Build VisualRoad benchmark image
git clone https://github.com/uwdb/visualroad
cd visualroad
docker build -t visualroad/core .
  • Install docker-compose Follow the instruction in this page

  • Generate Video!!

docker-compose run generator -s 1 -w 1920 -t 1080 -d 30 -r 1 test  # this is example

Modify the parameters based on your needs. Run docker-compose run generator -h to see all available options

About

Visual Road: A Video Data Management Benchmark

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 91.2%
  • Dockerfile 6.2%
  • Shell 2.6%