Docker

Docker is a service for managing containers, which you can think of as lightweight virtual machines. If you want to run Scanner in a distributed setting (e.g. on a cloud platform), Docker is essential for providing a consistent runtime environment on your worker machines, but it’s also useful for testing locally to avoid having to install all of Scanner’s dependencies. We provide prebuilt Docker images containing Scanner and all its dependencies (e.g. OpenCV, Caffe) at scannerresearch/scanner.

To start using Scanner with Docker, first install Docker. If you have a GPU and you’re running on Linux, you can install nvidia-docker (which provides GPU support inside Docker containers). Then run:

pip3 install --upgrade docker-compose
wget https://raw.githubusercontent.com/scanner-research/scanner/master/docker/docker-compose.yml
docker-compose run --service-ports cpu /bin/bash

If you installed nvidia-docker, then use gpu intead of cpu in the above docker-compose commands.

This installs the docker-compose utility which helps manage Docker containers. It uses the docker-compose.yml configuration file to create an instance of the Scanner docker image.

If these commands were successful, you should now have bash session inside the docker container. To start using Scanner to process videos, check out Getting Started.

The full set of docker configurations we provide are:

  • scannerresearch/scanner:cpu-VERSION - CPU-only build
  • scannerresearch/scanner:gpu-9.1-cudnn7-VERSION - CUDA 9.1, CUDNN 7
  • scannerresearch/scanner:gpu-8.0-cudnn7-VERSION - CUDA 8.0, CUDNN 7
  • scannerresearch/scanner:gpu-8.0-cudnn6-VERSION - CUDA 8.0, CUDNN 6

where VERSION is one of:

  • latest - The most recent build of the master branch
  • vX.X.X - A git tag (where X is an integer)